PATRICK E. HIGGINBOTHAM, Circuit Judge:
I
In April 1981, a number of independent Texas school districts sued National Gypsum Company, complaining of asbestos installed in approximately 600 school buildings located throughout the state of Texas. The School Districts sought recovery of $175 million and punitive damages of three times that amount for the costs of containing or replacing the asbestos materials and expenses incurred in taking other preventative measures to avert possible health hazards. These claims rested on several theories, including strict liability and negligence.
W.R. Grace & Co.-Conn.,1 a manufacturer of asbestos fireproofing material called Monokote, was added as a defendant in March 1984. In April 1984, Grace notified its insurers of the claim through its insurance broker, Marsh & McLennan, as required by the policies. In early 1987, Grace called a meeting of all its insurers to discuss the suit, including the progress of settlement negotiations. At this meeting Grace invited the insurance companies to participate in trial preparation and settlement negotiations and requested that the insurance companies pay for any settlement or judgment.
On April 14, 1987, Grace sought leave to file a third-party complaint against its primary insurer, Continental Casualty Company (CNA), and some eleven excess insurers.2 Grace had purchased primary liability insurance from CNA and excess liability insurance from each of the excess carriers during one or more of the years relevant to the School Districts’ claims. The third-party complaint sought protection from the School Districts’ claims on the theory that those claims alleged “property damage” resulting from an “occurrence” as those terms were defined in the policies of insurance. The district court granted leave to file the third-party complaint on April 20, 1987. On the same day the district court ordered the case bifurcated, with a separate trial for the third-party claim against the insurers.
Grace and the School Districts agreed to a tentative settlement on May 13, 1987, to be effective January 1, 1988. The agreement required approval by Grace’s management and by each School District. The district court then approved the settlement and discharged a jury seated for the trial of the School Districts’ claim against Grace.
In late May Grace sought summary judgment that CNA and the excess carriers were liable to Grace for its costs in the settlement. CNA and the excess carriers requested access to Grace’s documents. Grace's compliance with discovery requests is disputed.
On June 18, 1987, certain of the excess carriers moved to dismiss the third-party action for want of subject-matter jurisdiction. That same day one of the excess carriers requested that the third-party action be transferred for consolidation with a similar case involving Grace in New York,3 or to stay pending the decision in a case brought by Grace on May 26, 1987, against the excess carriers and others in Massachu[868]*868setts.4 The district court denied the motions to dismiss, transfer, or stay on September 3, 1987, and retained ancillary jurisdiction. Certain of the excess carriers filed a petition for mandamus, which this court accepted for review but then denied on October 2, 1987. In re AIU Ins. Co., No. 87-6024 (5th Cir. Oct. 2, 1987).
The settlement documents were formally executed on November 5, 1987, but the terms of the settlement were not disclosed to CNA and the excess carriers until January 15, 1988. The suit against Grace was dismissed on February 4, 1988.
II
CNA sold Grace standard form comprehensive general liability insurance as primary coverage for the period from July 1, 1978, to the present. The excess carriers sold Grace excess liability insurance for periods from July 1, 1978, through June 30, 1984.
CNA’s policies provide coverage and payment for “all sums which the Insured shall become legally obligated to pay as damages ... because of ... property damage ... caused by an occurrence_ [T]he Company shall have the right and duty to defend any suit against the Insured seeking damages on account of such ... property damage.... ” The term “occurrence” is defined as “an accident, event or continuous or repeated exposure to conditions which unintentionally causes injury to or destruction of property.” The term property damage is defined as “injury or destruction of property which occurs during the policy period, including the loss of use thereof at any time resulting therefrom.”
Similarly, the excess liability insurance policies generally provide coverage for “all sums which the Assured shall be obligated to pay by reason of the liability ... for damages on account of ... property damage ... caused by or arising out of each occurrence.... ” These policies generally define the term “occurrence” as “an accident or a happening or event or a continuous or repeated exposure to conditions which unexpectedly and unintentionally results in ... property damage ... during the policy period.” These excess policies, unlike standard primary policies, provided that the excess insurers had no duty to defend: “The Underwriters shall not be called upon to assume charge of the settlement or defense of any claim made or suit brought or proceeding instituted against the Assured.” Finally, the excess insurance companies’ policies generally define the term “ultimate net loss” to provide payment for:
[T]he total sum which the Assured ... becomes obligated to pay by reason of ... property damage ... either through adjudication or compromise, and shall also include ... all sums paid as ... fees ... and law costs ... expenses for ... lawyers ... and for litigation, settlement ... and investigation of claims and suits which are paid as a consequence of any occurrence covered hereunder.
The district court granted Grace’s motion for summary judgment in a memorandum opinion filed on February 4, 1988. Dayton Independent School Dist. v. National Gypsum Co., 682 F.Supp. 1403 (E.D.Tex.1988). The court ruled that:
(1) It was not required to make a choice of law because “the basic principles of policy construction in each jurisdiction are the same.” Id. at 1406.
(2) Grace need not “prove its actual liability as a prerequisite to obtaining coverage,” id., because “the only question ... is whether the claim asserted by [the school districts] describes a type of claim covered by the policies.” Id. at 1407.
(3) The complaint in the main action alleges “property damage” within the meaning of Grace’s policies. Id. at 1407-08.
(4) There are no genuine issues of material fact concerning whether Grace ei[869]*869ther expected or intended to cause property damage. Id. at 1408.
(5) Notwithstanding that different Excess Carriers were on the risk for different years, “[b]ecause the [School Districts] have claimed [continuous] property damage taking place from the installation of Grace’s products in the ... school buildings through the time those products were removed or contained, each policy at issue in this litigation is ... triggered to provide coverage to Grace.” Id. at 1410.
(6) “Grace may select which of its policies it wishes to provide coverage ... [and] may designate the policy year that will cover the claim in the Main Ac-tion_” Id. at 1411.
(7) None of the policy exclusions “bars coverage for [the School Districts’] claims or raises issues of fact which would preclude summary judgment.” Id. at 1412.
(8) Grace provided adequate notice of the claims against it. Id. at 1413.
(9) There are no genuine issues of material fact concerning whether Grace made misrepresentations or failed to disclose material facts to the excess carriers in its applications for insurance. Id. at 1414.
(10) The excess carriers waived their defenses of late notice and misrepresentation or concealment. Id. at 1414.
With its memorandum the court filed an order under seal granting summary judgment against all of the excess carriers except Gerling, which had not yet been served. Summary judgment was entered against Gerling on May 2, 1988.
On June 3, 1988, Grace moved for the entry of a final judgment that: (a) assigned the settlement amount to the excess carriers who sold policies during the 1981-1982 policy period; (b) ordered the primary insurer, CNA, to pay Grace’s fees and costs incurred; and (c) awarded Grace its attorneys’ fees in prosecuting the third-party action under Article 38 of the Texas Civil Practice and Remedies Code. CNA did not contest that portion of Grace’s motion for the payment of fees and costs in the main action. On June 27 Grace also moved for prejudgment interest under Tex.Rev.Civ. Stat.Ann. Art. 5069-1.05 (Vernon Supp.1985).
At a July 6, 1988 hearing, the district court granted Grace’s motion for final judgment, except it deferred Grace’s motion for counsel fees in the third-party action. Final judgment with prejudgment interest was entered on August 23, 1988. The judgment required six third-level excess carriers to “drop down” and fill the gap in coverage caused by the insolvency of one of Grace’s second level excess insurers. Also, judgment for fees and costs in the main action was not reduced by the $500,000 per occurrence deductible. An amended final judgment was entered on September 8, 1988, enlarging the award of prejudgment interest.
After discovery concerning the reasonableness of Grace’s attorneys’ fees in the third-party action, Grace moved for a supplemental judgment on that issue. On November 7, 1988, the court ordered all the insurance companies to indemnify Grace for attorneys’ fees and costs incurred in the third-party action. In addition, the court ruled that Grace would be entitled to recover the attorneys’ fees and costs that it incurred in connection with this appeal.
Ill
Grace’s third-party claim against its insurer had no source of federal jurisdiction independent of the primary suit. The sole basis for jurisdiction over Grace’s claims against the insurers was that they were ancillary to the School Districts' claims against Grace. The primary suit was in federal court because of the diversity of citizenship of the parties. In May 1987, after the tentative settlement of the School Districts’ claims against Grace, the district court decided to retain jurisdiction over Grace’s claims against its insurers.5 The district court did not stay any action per-[870]*870taming to the third-party suit, but instead pushed forward toward its resolution, despite the pending settlement. The insurers argue that the district court abused its discretion in refusing to dismiss Grace’s third-party claims against them for lack of jurisdiction. There had never been any federal claims involved, and once the primary suit was settled, there was no longer complete diversity of citizenship, as Grace and several of the insurers were citizens of New York.
We review the district court’s continuing exercise of ancillary jurisdiction by an abuse of discretion standard. United States v. City of Twin Falls, Idaho, 806 F.2d 862, 868 (9th Cir.1986), cert. denied, 482 U.S. 914, 107 S.Ct. 3185, 96 L.Ed.2d 674 (1987). See also Joiner v. Diamond M Drilling Co., 677 F.2d 1035, 1042 (5th Cir.1982). By this measure we affirm absent a definite and firm conviction of clear error. We are convinced that the district court abused its discretion in retaining ancillary jurisdiction here.
Although a tentative settlement in the main action was announced in May 1987, the settlement was not concluded until November 1987, and the supporting claim against Grace was not dismissed until February 1988. Thus, the basis for the district court’s original—and clearly proper—subject-matter jurisdiction over the third-party action arguably remained until February 1988. This assumes, of course, that dismissal was not delayed in order to manufacture ancillary jurisdiction. The insurers hint darkly at just that. Indeed, a reasonable mind might be suspicious. Nonetheless, we are left with only little more than objective facts in the record, and we must therefore, consider whether the district court’s retention of ancillary jurisdiction was an abuse of discretion given the situation in February 1988.
The premier case in this circuit addressing the exercise of ancillary jurisdiction after the settlement of the main suit is Joiner v. Diamond M. Drilling Co., 677 F.2d 1035 (5th Cir.1982). In Joiner we noted that “[generally, when the primary federal claim has been settled or dismissed before trial, the district court should dismiss any lingering ancillary state law claims.” Id. at 1041 (citing United Mine Workers v. Gibbs, 383 U.S. 715, 726, 86 S.Ct. 1130, 1139, 16 L.Ed.2d 218 (1966) (“[I]f the federal claims are dismissed before trial, the state claims should be dismissed as well.”)). See also Waste Systems Inc. v. Clean Land, Air, Water Corp., 683 F.2d 927, 930 (5th Cir.1982); Putnam v. Williams, 652 F.2d 497, 502 (5th Cir.1981) (“... the preferred practice, when no extensive proceedings on the ancillary claim have begun in federal court, is to dismiss [the state-law claim] so that the case may be brought in state court where it belongs.”); McDonald v. Oliver, 642 F.2d 169, 172 (5th Cir.1981) (“... the district court lost its ancillary jurisdiction when the underlying lawsuit was dismissed.”); Tinker v. DeMaria Porsche-Audi, Inc., 632 F.2d 520, 523 (5th Cir.1980) (“[I]f the federal claims are dismissed before trial ... the state claim should be dismissed as well.”).
We recognized, however, that there would be circumstances when the exercise of jurisdiction over the third-party claim could be proper. “While Gibbs suggests that dismissal of appended state law claims is mandatory whenever the main federal claim is dropped prior to trial, in practice district courts have been allowed a measure of discretion in determining whether to retain their ancillary jurisdiction over appended state claims.” Joiner, 677 F.2d at 1042. Several factors must be considered whenever a federal court adjudicates pendent or ancillary claims in order to determine whether this exercise of jurisdiction at the outer limits of federal power has been an abuse of discretion. As we stated in Joiner:
Gibbs reminds us that judicial efficiency and economy is the raison d’etre of ancillary jurisdiction. If the interests of judicial economy are no longer advanced by the retention of ancillary claims, the state-law action ought to be dismissed.
Gibbs also tells us that federal courts must be ever mindful of the states’ interest in the integrity and autonomous development of their own jurisprudence.
[871]*871“Needless decisions of state law should be avoided ... as a matter of comity”, and “[fjederal courts should not be overeager to hold on to the determination of issues that might be more appropriately left to settlement in state court.” Thus, Gibbs teaches that federal courts must refrain from unnecessary poaching upon a sovereign state’s jurisprudential turf.
677 F.2d at 1042 (citations omitted).
The Supreme Court has recently stated that a federal court may not assert authority over an ancillary claim solely because the Gibbs test for pendent jurisdiction has been met. Finley v. United States, - U.S. -, 109 S.Ct. 2003, 2008, 104 L.Ed.2d 593 (1989). The Court referred to its earlier opinion in Owen Equipment & Erection Co. v. Kroger, 437 U.S. 365, 98 S.Ct. 2396, 57 L.Ed.2d 274 (1978), where it held “neither the convenience of the litigants nor considerations of judicial economy can suffice to justify extension of the doctrine of ancillary jurisdiction.” Finley, 109 S.Ct. at 2008 (citing Kroger 437 U.S. at 376-77, 98 S.Ct. at 2403-04).
The ancillary jurisdiction of the federal courts was originally recognized in cases such as Freeman v. Howe, 24 How. 450, 16 L.Ed. 749 (1861), which held that when federal jurisdiction “effectively controls the property or fund under dispute, other claimants thereto should be allowed to intervene in order to protect their interests, without regard to jurisdiction.” Aldinger v. Howard, 427 U.S. 1, 11, 96 S.Ct. 2413, 2419, 49 L.Ed.2d 276 (1976). Although ancillary jurisdiction has been expanded to include cases such as compulsory counterclaims,6 impleader,7 cross-claims,8 and third-party claims,9 “the context in which the non-federal claim is asserted is crucial.” Owen Equipment & Erection Co. v. Kroger, 437 U.S. 365, 376, 98 S.Ct. 2396, 2403, 57 L.Ed.2d 274 (1987). Ancillary jurisdiction involves “claims by a defending party haled into court against his will”, or by a party “whose rights might be irretrievably lost unless he could assert them in an ongoing action in a federal court.” Id.
Ancillary jurisdiction involves a more intrusive assumption of power [than pendent jurisdiction]. Because ancillary jurisdiction extends jurisdiction to additional parties rather than only additional claims between the original parties, it potentially encroaches more heavily on state autonomy and the rights of the parties. The exercise of ancillary jurisdiction is therefore a greater assumption of power. The courts must make a more searching inquiry to ascertain if Congress has granted this power. The fairness concerns raised by the party asserting the ancillary claim must outweigh the negative impact on state autonomy and the ancillary party.
Note, Unraveling the “Pendent Party” Controversy: A Revisionist Approach to Pendent and Ancillary Jurisdiction, 64 B.U.L. Rev. 895, 929 (1985).
“While there is no ancillary jurisdiction paradigm, nevertheless, there is a common threat in all ancillary jurisdiction cases; the party asserting a non-federal claim may effectively lose that claim if prevented from asserting it in the federal court as part of the federal case.” Id. at 907. It is fairness to a party who has not instigated the litigation in federal court, whose rights might be lost, that is the primary reason for permitting ancillary jurisdiction. “Ancillary jurisdiction exists because without it the federal court neither could dispose of the principal case effectively nor do complete justice in the dispute that is before the tribunal.” 13 C. Wright, A. Miller & E. Cooper, Federal Practice & Procedure § 3523, p. 85 (1984 ed.) (citing Rogers v. Aetna Cas. & Sur. Co., 601 F.2d 840 (5th Cir.1979)).
[872]*872When the underlying suit is settled, the rights involved in the third-party action and the main action are no longer tied together. The federal court’s ability to effectively dispose of the principal case no longer depends upon its exercise of jurisdiction over the third party action. Provided there is no prejudice to the former defendant and third-party plaintiff, there is no reason to keep an ancillary claim in federal court after the main action has settled, unless the settlement of the main action effectively disposes of the third-party action. There is no longer the threat that the party “haled into federal court” will lose its third-party claims unless they are adjudicated in federal court.
Once the main action has reached settlement prior to trial, there is generally little that can be gained in the way of judicial efficiency and economy by continuing to adjudicate ancillary state-law claims, and considerations of comity and federalism militate in favor of dismissal. A federal court may retain jurisdiction over ancillary state-law claims, however, if dismissal after a pretrial settlement of the main action would unduly prejudice the parties. There was no such prejudice demonstrated here. As in a similar case where this court held that the continuing exercise of ancillary jurisdiction was an abuse of discretion, “no trial has taken place, [and] any discovery could be utilized in a state court proceeding.” Waste Systems v. Clean Land, Air, Water Corp., 683 F.2d 927, 931 (5th Cir.1982).10 Grace has not shown that the applicable statute of limitations would have run while the state claim was pending in federal court, thereby depriving Grace of any forum in which to bring its claims. On the contrary, the same claims are pending in several courts around the country, and Grace can still sue the insurers in New York state court. Nor had there been substantial commitment of judicial resources to the nonfederal claims at the time that settlement was reached.
Grace argues that a substantial amount of activity had already occurred in the federal court relating to the ancillary claims by the time the main action against Grace was dismissed, and that judicial efficiency and economy would be served by the district court’s retention of jurisdiction over the third-party action because the district court was familiar with the suit, had presided over the primary suit for a long time prior to the dismissal, and was familiar with the applicable law and facts. The suit was in federal district court in the Eastern District of Texas, and the judge was very familiar with Texas law. Therefore, Grace argues, judicial efficiency would be advanced by the court's continuing exercise of ancillary jurisdiction.
The error of these arguments is manifest once the district court’s choice of law is examined. As we will demonstrate, the district court’s familiarity with the main action and with Texas law did not make its exercise of jurisdiction efficient in this case, nor was there any reason for the case to be decided in a federal forum located in Texas, for neither Texas law nor general rules of contract construction were the correct law to be applied to the third-party suit. The Texas interests dropped out when the primary action settled. The School Districts were paid, and the action was reduced to a dispute between a New York company and its New York insurers, where the governing law would be the law of New York. The factors of judicial efficiency and economy and concerns with federalism and comity all militated for dismissal of the third-party action so that “substantial novel questions of state law” could be decided in state court. Jones v. Fitch, 665 F.2d 586, 593 (5th Cir.1982).
IV
We now turn to a discussion of the proper application of state law and the interests that the states of Texas and New York have in this case. The excess carriers asked the district court to apply New York [873]*873law.11 The district court refused to do so, reasoning that it was not required to make a choice of applicable law because “the basic principles of policy construction in each jurisdiction are the same.” 682 F.Supp. at 1406. This was error. As will be seen, the law of New York is both fairly well developed and different from the law of Texas. Also, in those areas where the law of New York is unclear, it could arguably be resolved differently from Texas, and it is certainly in the interest of New York for New York courts to determine its law. Where the law of different jurisdictions is in direct conflict, or where the law is unclear, Erie R.R. Co. v. Tompkins, 304 U.S. 64, 58 S.Ct. 817, 82 L.Ed. 1188 (1938), dictates that the district court apply state law, not some general law of construction, and in order to do so the district court must choose the law of one state that is most properly applied to the dispute before it.
The Texas law of conflicts, the state in which the district court sits, governs the substantive issues here. See Klaxon Co. v. Stentor Electric Manufacturing Co., 313 U.S. 487, 61 S.Ct. 1020, 85 L.Ed. 1477 (1941); NCH Corp. v. Broyles, 749 F.2d 247 (5th Cir.1985); Stuart v. Spademan, 772 F.2d 1185, 1195 (5th Cir.1985). Under Texas choice-of-law principles, contract disputes are governed by “the law of the state with the most significant relationship to the particular substantive issue.” Duncan v. Cessna Aircraft Co., 665 S.W.2d 414, 421 (Tex.1984). Our issues revolve around the construction and application of insurance policies.
New York has the most significant relationship to the substantive issues in this case. Grace and three of the insurers maintain their principal places of business in New York. At least one other insurer also has an office in New York, and many of the insurers do business there. Grace’s insurance broker is based in New York and most of the policies were solicited, negotiated, and delivered in New York. Grace paid its premiums in New York, and gave its notice of the school district’s claims in New York. New York was, by all accounts, the center of the relationship between Grace and the insurers. New York has a strong and overriding interest in the outcome of insurance coverage disputes, such as this, which involve a New York insured and its contractual relationships with insurers doing business in New York.
Grace suggests that Texas rather than New York law should control because this will determine the source of funds for payment of its settlement with the School Districts. We rejected a similar argument in Atlantic Mutual Insurance Co. v. Truck Insurance Exchange, 797 F.2d 1288 (5th Cir.1986). Atlantic Mutual arose out of injury allegedly caused by the insured, Mr. Santini, in Texas. The claim was settled by Santini and one of its carriers, Atlantic Mutual Insurance Co., who then sued a second carrier, Truck Insurance Exchange, for contribution. As Grace does here, Truck argued that because “settlement of the [underlying] lawsuit was effected in Texas based on an injury that occurred in Texas,” Texas law should apply. Id. at 1292. We held, however, that under the applicable Texas choice-of-law rules, New York law governed.
We noted that both Santini and Atlantic maintained their principal places of business in New York and that “New York has an interest not only in Santini’s recovery, but also in the application of its insurance laws to Atlantic’s claim against Truck.” Id. at 1291-92. We further observed that “the expectations of [Santini, Atlantic, and Truck], which we believe point to the application of New York law as to each policy, are more important than the expectations between [the injured claimant] and Santi-ni.” Id. at 1292. Because the policies provided nationwide liability coverage, the fact that Santini was based in New York “point[ed] with particular strength to the application of New York law.” Id. We therefore concluded that, “[although Texas has a strong interest in [the injured claimant’s] recovery against Santini, New York has a more significant relationship to [874]*874the questions that determine Atlantic’s recovery.” Id.
It is important that neither Texas nor the School Districts have any interest in whether the settlement is paid by Grace or, instead, by its insurers. Their only interest is in being paid, and they have been. Texas’s interest in the School Districts’ recovery against Grace ended with its settlement and payment. What remained was a dispute between Grace and its insurance carriers. It is New York, not Texas, that has the greater interest in and most significant relationship to the questions of insurance law that decide this dispute.
V
Grace argues that there is no need to apply New York law, for the district court was correct in its finding that the laws of all of the relevant jurisdictions are the same on the important issues in this case. If the laws of the states do not conflict, then no choice-of-law analysis is necessary. Keene Corp. v. Insurance Co. of North America, 667 F.2d 1034, 1041 (D.C.Cir.1981), cert. denied, 455 U.S. 1007, 102 S.Ct. 1644, 71 L.Ed.2d 875 (1982). Grace contends that there is no conflict between the laws of Texas and the laws of New York, so there is no reason to choose between them. This contention is not correct; although there are areas where the laws of the two states are in harmony, there are significant differences in the laws of the two states on some issues. Moreover, there are important areas where either the law of New York or the law of Texas is not clear.
In addition, the district court applied general rules of construction favoring the insured, without first applying the substantive decisional law of either New York or any specific jurisdiction. This was error. General rules of construction come into play “only as a matter of last resort.” Schering Corp. v. Home Ins. Co., 712 F.2d 4, 10 n. 2 (2d Cir.1983). See also Bybee v. John Hancock Mut. Life Ins. Co., 507 S.W.2d 330, 331 (Tex.Civ.App.-Tyler 1974, writ ref’d n.r.e.). Of course, in situations where the substantive decisional law of all relevant jurisdictions is the same, a court need not go through the motions of making a choice of law. But this is not such a case. On several important issues of insurance law presented by this case the law of New York is either different from the “general rules” applied by the district court, or is unclear. These issues would best be left to be decided in state court.
One important issue of insurance law in this case is whether an insurer is liable to indemnify its policyholder for sums paid in settlement, where the allegations of the underlying claim may be within the policy coverage but no underlying facts have been proved. The district court held that the nature of the third party’s claim is determinative, and not the actual facts underlying that claim. The insurers argue that this is contrary to New York law, which holds that there is no duty to indemnify unless settled claims are actually, not merely allegedly, covered by the policy of insurance. Servidone Construction Corp. v. Security Insurance Co., 64 N.Y.2d 419, 488 N.Y.S.2d 139, 477 N.E.2d 441 (1985). In Servi-done the insured had settled a claim after the insurer had breached its duty to defend. The insured then sued the insurer to recover the amount of the settlement, and the insurer argued that the loss was not covered. The New York Court of Appeals reversed the district court’s imposition of liability, reasoning that “an insurer’s breach of duty to defend does not create coverage and that, even in cases of negotiated settlements, there can be no duty to indemnify unless there is first a covered loss.” 64 N.Y.2d at 423, 488 N.Y.S.2d 139, 477 N.E.2d 441.
The duty to defend is measured against the allegations of pleadings but the duty to pay is determined by the actual basis for the insured’s liability to a third person .... By holding the insurer liable to indemnify on the mere “possibility” of coverage perceived from the face of the complaint — the standard applicable to the duty to defend — the court has enlarged the bargained for coverage as a penalty for the breach of the duty to defend, and this it cannot do.
[875]*875Id. at 424, 488 N.Y.S.2d 139, 477 N.E.2d 441 (citations omitted). See also Apache Foam Products v. Continental Ins. Co., 139 A.D.2d 933, 528 N.Y.S.2d 448 (1988). This law of the state of New York is plainly different from the law applied by the district court, and shows a conflict of laws necessitating a choice of law decision. As we have already shown, the choice of law analysis shows that the state with the most significant relationship to the dispute on this issue is the state of New York.
Another important issue in this case is the meaning of “property damage” for purposes of imposing insurance coverage. Although the School Districts sued with a theory that the asbestos in the fireproofing materials they purchased from Grace caused “property damage,” it is unclear whether there was actual property damage so as to require coverage under the insurance policies. Grace’s policies defined “property damage” as “loss of or direct damage to or destruction of tangible property.” The School Districts’ claims rested on the presence and incorporation of Grace’s product into the buildings, and allegations that the asbestos fibers were being released from the fireproofing materials of which they were a part, and were being impregnated into the other materials in the building. Grace and the district court relied on the School Districts’ allegation that ceilings, walls, floors, etc. had become dusted with asbestos particles. It is not clear that such dusting could be considered “property damage” under the terms of the policy and New York law. The real harm alleged by the School Districts concerned the need to remove or contain the asbestos building materials themselves. It is not clear that costs associated with such removal should be treated as property damage. Courts that have considered the issue have disagreed concerning whether the cost of removal should be treated as covered “property damage” or noncovered economic loss. See, e.g., Maryland Casualty Co. v. Armco, Inc., 822 F.2d 1348, 1353 (4th Cir.1987), cert. denied, 484 U.S. 1008, 108 S.Ct. 703, 98 L.Ed.2d 654 (1988) (costs incurred in taking preventative measures to avert possible future health risks do not represent current “property damage”); Millers Mutual Fire Ins. Co. v. Ed Bailey, Inc., 103 Idaho 377, 647 P.2d 1249, 1253 (1982) (concern over the possible need for future containment or removal does not amount to property damaged during past policy periods); Hamilton Die Cast, Inc. v. United States Fidelity & Gaurantee Co., 508 F.2d 417, 419-20 (7th Cir.1975) (“[T]he mere inclusion of a defective component, where no physical harm to the other parts results therefrom, [does not] constitute[] ‘property damage’ within the meaning of the policy.”). But, cf., Bowman Steel Corp. v. Lumbermens Mutual Casualty Co., 364 F.2d 246, 249 (3d Cir.1966); Pittsburgh Plate Glass Co. v. Fidelity & Casualty Co., 281 F.2d 538, 541 (3d Cir.1960). The courts of New York have not had occasion to decide how the term “property damage” applies in cases such as this. Considerations of comity and federalism, therefore, suggest that retention of ancillary jurisdiction was improper and that this issue of state law should have been left for resolution by a state court.
The district court also concluded that: Because the [School Districts] have claimed property damage taking place from the installation of Grace’s products in the ... school buildings through the time those products were removed or contained, each policy at issue in this litigation is, therefore, triggered to provide coverage to Grace.
682 F.Supp. 1403, 1410.
Grace’s policies provided coverage only when property damage resulted “during the policy period.” Under New York law, such policy language is construed, in accordance with its plain meaning, to restrict coverage to property damage that in fact occurs during thé policy period. In American Home Products Corp. v. Liberty Mutual Ins. Co., 565 F.Supp. 1485 (S.D.N.Y.1983), the district court, applying New York law, held that policies restricting coverage in the same manner were triggered only upon “a showing of actual injury ... occurring during the policy period, based upon the facts proved in each particular case.” Id. at 1489, 1497. The Court of [876]*876Appeals affirmed. 748 F.2d 760 (2d Cir.1984). American Home Products was a personal injury case, but the same reasoning was applied to find that the same language required injury in fact to trigger coverage in a case involving claims of both personal injury and property damage. Olin Corp. v. Insurance Co. of North America, 603 F.Supp. 445, 448 (S.D.N.Y.1985). Uniroyal, Inc. v. Home Ins. Co., 707 F.Supp. 1368 (E.D.N.Y.1988), is not to the contrary. Judge Weinstein expressly held that “New York law requires that injury in fact theory under a comprehensive general liability policy. Under that theory, ‘coverage [is] based upon the occurrence during the policy period of an injury in fact.’ ” Id. at 1388 (quoting American Home Products, 748 F.2d at 764). This is different from the continuous trigger theory applied by the district court here. All three of these cases were in federal court, but the courts attempted to apply New York law. The district court characterized two other decisions applying New York law as “triggering successive policies in effect during a continuous course of injury.” But those decisions recognized that coverage was triggered only by “the sustaining of a specified injury during the policy period.” National Casualty Ins. Co. v. City of Mount Vernon, 128 A.D.2d 332, 515 N.Y.S.2d 267, 270 (1987). See also McGroarty v. Great American Insurance Co., 36 N.Y.2d 358, 361, 368 N.Y.S.2d 485, 329 N.E.2d 172 (Ct.App.1975).
New York and Texas appear to disagree with respect to the trigger of coverage. Grace cites National Standard Ins. Co. v. Continental Ins. Co., No. CA-3-81-1015-D (N.D.Tex. Oct. 4, 1983), for the proposition that Texas applies a “continuance trigger,” at least for the duty to defend certain bodily injury claims. There appears to be no Texas case applying either the injury-in-fact standard or a continuous trigger to the duty to indemnify in property damages cases. Thus, either Texas law is unclear, or it would conflict with New York law.
The amici argue that personal injury cases do not state the law for property damage cases, and that neither the “continuous trigger” nor the “injury in fact” tests are proper, but that the “first discovery” principle should be applied. They argue that property damage is fundamentally an economic matter, and that damage to property cannot occur in any meaningful sense until the damage is perceived. In American Motorist Insurance Co. v. Levelor Lorentzen, Inc., No. 88-1994, 1988 WL 112142 (D. N.J., October 14, 1988), the court, looking to new York law, held that: “In the context of damage to property caused by contamination from hazardous waste the ‘first discovery’ principle is the most appropriate standard to apply....” Slip op. at 9-10. The amici assert that asbestos in buildings claims are no different from other claims of alleged latent, continuing, or progressive property damage. It is clear that general principles of contract construction cannot answer the question of what will trigger coverage, and it is also clear that the law of New York either squarely conflicts with that applied by the district court, or at least is not settled. In either case, for reasons of federalism and comity, the district court should have declined to exercise ancillary jurisdiction so this important issue of New York law could be decided by a New York court.
The excess carriers also argue that New York law differs from the general rules applied by the district court on the issue of the excess carriers defense of nondisclosure. “[U]nder New York law, even an innocent misrepresentation is sufficient to allow an insurer to avoid a contract.” Friedman v. Prudential Life Ins. Co., 589 F.Supp. 1017, 1022 (S.D.N.Y.1984). The district court applied Texas law that allows a defense only in the case of a “misrepresentation ... made willfully with the intent to deceive.” Lee v. National Life Assurance Co., 632 F.2d 524, 527 (5th Cir.1980). At least one New York case has held that damage is unintended or unexpected only if it is not reasonably foreseeable. Drew Chemical Corp. v. Fidelity and Casualty Co. of New York, 96 Misc.2d 503, 413 N.Y.S.2d 538 (Sup.Ct.1976), aff'd, 60 A.D.2d 552, 400 N.Y.S.2d 334 (1977), aff'd, 46 [877]*877N.Y.2d 851, 414 N.Y.S.2d 515, 387 N.E.2d 226 (1979). The rule in New York appears to be that an insured may still be covered if it engages in conduct which it knows might cause damage, but not if it engages in conduct which it knows or should know is likely to do so. The law of New York is therefore either in conflict with that applied by the district court, or arguably unclear. In either case this dispute indicates that it was an abuse of discretion for the district court to continue to exercise ancillary jurisdiction.
VI
The district court abused its discretion in retaining ancillary jurisdiction over these claims, for they involve hotly disputed issues that are important to the development of New York insurance law. The district court erred in its application of Texas law or general principles of contract construction, for New York is the state with the most significant relationships to the dispute, and there are conflicts between New York law and the law as applied by the district court. REVERSE with Order to Dismiss for lack of jurisdiction.