DYK, Circuit Judge.
Plaintiff-Appellant Howmedica Osteonics Corporation (“Howmedica”) appeals from a final judgment of noninfringement in this patent infringement action against defendant-appellee Wright Medical Technology, Inc. (“Wright”). Under the district court’s claim construction of the term “femoral component including at least one condylar element,” the parties stipulated that the accused product did not infringe the asserted patent claims. Because we conclude that the construction of this claim term was incorrect, we cannot sustain the stipulated judgment. Furthermore, because we agree with the district court that Howmedica did not release its infringement claim as part of an earlier settlement agreement, Wright’s asserted alternative ground for affirmance does not support the judgment. Accordingly, we vacate the judgment of noninfringement, and remand to the district court for further proceedings consistent with this opinion.
[1340]*1340BACKGROUND
I
Both parties develop, manufacture, and sell orthopedic implants for use in the reconstruction of various joints of the human body. Howmedica is the owner of U.S. Patent No. 5,824,100 (“the '100 patent”), which is directed to an artificial knee prosthesis or implant used to replace part or all of a patient’s knee joint. Claim 15, the only independent claim asserted in this action, provides in full:
In a knee prosthesis for replacing the natural knee, the knee prosthesis having a femoral component and a tibial component, the tibial component including a bearing member and the femoral component including at least one condylar element for confronting and engaging the bearing member to accomplish articulation of the knee prosthesis throughout a range of flexion, including a primary range of flexion between a hyperextended position and a flexed position, the engagement between the condylar element of the femoral component and the bearing member of the tibial component ordinarily taking place at a contact area along articular surface areas of the condylar element and the bearing member, the improvement comprising:
anterior-posterior surface profile contours along the condylar element and the bearing member, the anterior-posterior surface profile contour along the condylar element having an essentially constant anterior-posterior articular radius throughout the articular surface area of the condylar element which contacts the bearing member during articulation throughout the primary range of flexion, the anterior-posterior articular radius having an origin lying generally along a line extending laterally between the medial and lateral collateral ligament attachment points on the femur of the natural knee.
'100 patent col. 6 1.58-col. 8 1.3 (emphases added). Claim 18, which depends from claim 15, is also at issue in this litigation. Claim 18 describes: “The improvement of claim 15 wherein the hyperextended position is at about -15° in the range of flexion, and the flexed position is at about 75° in the range of flexion.” Id. col. 8 11.8-10.
The knee joint is the point of connection between the lower end of the femur (thigh bone) and the upper end of the tibia (shin bone). In a natural knee, the lower end of the femur includes two rounded, generally smooth projections of bone called condyles, which engage with the upper end of the tibia. As pictured below, the femoral component (22) of the prosthesis described in claim 15 replaces a portion of the natural femur, including one or both condyles. A prosthesis is called “unicondylar” if it replaces only one condyle of the natural femur, and “bieondylar” if it replaces both natural condyles. The condylar element or elements (54) of the femoral component articulate against a “bearing member” (64), part of the “tibial component” (24) of the prosthesis, which is surgically implanted to replace an upper portion of the tibia.
The constant-radius geometry of the condylar element described by claim 15 furthers several goals: it allows “increased areas of contact between the condylar elements” and the tibial components “for lowered stress in the material of the bearing member during articulation of the knee prosthesis”; it facilitates “a better balance of the tension in the collateral ligaments of the knee”; it “enables a higher degree of conformity between the condylar elements and the bearing member for reduced stresses during normal activity”; it simplifies the configuration of the components, allowing easier manufacturing; and it “en[1341]*1341ables exemplary performance and increased reliability over an extended service life.” Id. col. 111.48-62.
[[Image here]]
II
The '100 patent issued on October 20, 1998. Howmedica did not immediately bring suit against Wright on the '100 patent. However, Howmedica in 1997 had commenced two lawsuits in New Jersey against Wright alleging infringement of U.S. Patent Nos. 5,192,324 (“the Kenna patent”) and 5,133,772 (“the Hack patent”). In 1999, Howmedica filed a third lawsuit against Wright in New Jersey, alleging infringement of U.S. Patent No. 4,892,549 (“the Figgie patent”). At the same time, a lawsuit filed in 1991 by Wright’s predecessor against Howmedica, alleging infringement of U.S. Patent No. 4,474, 177 (“the Whiteside patent”), was pending in Massachusetts.
In April 1999 the parties initiated settlement negotiations directed to resolving the four pending actions. These negotiations took place between Thomas Patton (Wright’s President and Chief Executive Officer), Ned Lipes (Howmedica’s President), Alfred Zarnowski (Howmedica’s Senior Director of Intellectual Property), and the two parties’ outside patent counsel. The parties reached an agreement at a meeting on November 15, 1999. Wright’s counsel was responsible for drafting a written agreement reflecting the negotiations. At this point the parties contemplated that they would execute a single written agreement.
On November 23, 1999, Wright’s counsel explained to Howmedica’s counsel that they had “broken the settlement into two separate agreements since Wright Medical says this is required to comply with generally accepted accounting practices.” J.A. at 978. One agreement, the New Jersey agreement, covered the Kenna, Hack, and Figgie patent lawsuits. The second agreement, the Massachusetts agreement,, covered the Whiteside patent lawsuit. The two draft agreements sent to Howmedica [1342]*1342by Wright included nearly identical mutual releases, which stated:
Howmedica Osteonics hereby releases and forever discharges Wright Medical ... from any and all manner of claims, controversies, demands, damages, causes of action, or suits, that Howmedica Osteonics has, have had, or may have against Wright Medical ... upon or by reason of or relating to any acts or omissions made by Wright Medical ... on or before the effective date of this Agreement, including, but not limited to, any and all claims and counterclaims that were or could have been asserted by Howmedica Osteonics in the lawsuit.
J.A. at 574 (draft Massachusetts agreement) (emphases added); see also J.A. at 586 (draft New Jersey agreement containing substantively identical language).1
In response to the draft agreements, Howmedica objected to the release provision of the Massachusetts agreement as being “too broad” and therefore not accurately reflecting the parties’ agreement. J.A. at 389. Howmedica explained that, “[a]s drafted, Howmedica Osteonics would be releasing and discharging Wright Medical from claims involving patents owned by Howmedica Osteonics and other claims having nothing to do with any of the matters currently in dispute between the parties.” Id. However, Howmedica did not separately object to the identical provision in the New Jersey agreement, although it did request other changes to that draft. In response to Howmedica’s objection, Wright’s counsel struck the words “including, but not limited to, any and all claims and counterclaims” from the release provision of the Massachusetts agreement. As revised, the Massachusetts release provision covered only claims or controversies “that were or could have been asserted” in the Whiteside patent lawsuit. See id. at 382. The “including, but not limited to” language was not stricken from the release provision of the New Jersey agreement, a redlined version of which was sent to Howmedica along with the revised Massachusetts agreement. The language of the New Jersey agreement was never conformed to the language of the Massachusetts agreement. Both settlement agreements were executed as of December 21, 1999.
Ill
On March 10, 2000, less than three months after the execution of the releases, Howmedica filed this' action against Wright, alleging infringement of the '100 patent by Wright’s ADVANCE® Total Knee System. Wright responded, asserting affirmative defenses of noninfringement and invalidity, and also filed a counterclaim seeking a declaration that the '100 patent was invalid and not infringed. Wright did not initially include an affirmative defense based on the release provision in the New Jersey agreement. However, after the case had been pending for over two years, Wright filed an amended answer and counterclaim, asserting as new affirmative defenses that the '100 patent was unenforceable for misuse and inequitable conduct, and that Howmedica’s action was barred by the New Jersey release provision.
Both parties subsequently cross-moved for summary judgment on Wright’s release defense. In the memorandum in support of its summary judgment motion, Howmedica argued that the New Jersey release provision, based on the intentions of the parties, covered only the patents [1343]*1343that were the subject matter of the four pending lawsuits resolved by the settlement. In opposition, Wright argued that the broad language of the New Jersey release provision barred the current action. Wright characterized Howmediea’s position as seeking reformation of the New Jersey release, and argued that reformation was not appropriate because there was no mutual mistake as to the language of the provision. In reply, Howmedica urged that reformation was unnecessary under the New Jersey law of contract interpretation. The parties also took these positions in the simultaneous briefing on Wright’s cross-motion for summary judgment.
On March 18, 2005, the district court denied Wright’s motion and granted summary judgment in favor of Howmedica on the release defense. The court treated the New Jersey and Massachusetts agreements as though they were a single agreement, noting that “[t]he releases encompassed within the agreements were executed as part of the same settlement or compromise of the then-pending patent lawsuits.” J.A. at 62. The court determined that the parties intended and understood the releases in the agreements to cover only the four lawsuits pending at the time (as well as one additional patent that was explicitly included in the New Jersey agreement). The court held that, under New Jersey law, it could interpret the provisions to not cover Howmedica’s claim under the '100 patent. The district court concluded that reformation of the agreement to reflect the intention of the parties was thus unnecessary.
On November 29, 2005, following briefing and a Markman hearing, the district court issued a claim construction order construing three disputed terms of claim 15 of the '100 patent. First, the court construed the phrase “femoral component including at least one condylar element” to require, in a bicondylar femoral component, both condyles to meet the geometric limitations of the claim. Second, the court construed “primary range of flexion” to be limited to a particular range of -15° of hyperextension to +75° of flexion, in accordance with the specification of the '100 patent. Finally, the court determined that the phrase “lying generally along a line extending laterally between the medial and lateral collateral ligament attachment points on the femur of the natural knee” was not indefinite.
Howmedica acknowledges that it could not prove that both condylar elements of Wright’s accused products meet the geometric limitations of claim 15 throughout the primary range of flexion and, therefore, that it could not prove infringement under the district court’s construction of the term “femoral component including at least one condylar element.” Accordingly, Howmedica stipulated to a final judgment of noninfringement. Howmedica timely appealed that judgment to this court.2 We have jurisdiction pursuant to 28 U.S.C. § 1295(a)(1).3
[1344]*1344DISCUSSION
We first address claim construction. Claim construction is an issue of law that we review de novo. See Cybor Corp. v. FAS Techs., Inc., 138 F.3d 1448, 1456 (Fed.Cir.1998) (en banc). On appeal, Howmedica challenges the district court’s construction of the term “femoral component including at least one condylar element.”
Claim 15 of the '100 patent recites a knee prosthesis with a “femoral component including at least one condylar element” and requires that the condylar element have a certain geometry, namely, that “the anterior-posterior surface profile contour” have “an essentially constant anterior-posterior articular radius throughout the articular surface area of the condylar element which contacts the bearing member during articulation throughout the primary range of flexion.” 100 patent col. 6 11.61-62, col. 7 11.5-11. It is undisputed that claim 15 can cover a unicondylar prosthesis, where only one artificial condyle is employed and one natural condyle remains. However, based on the use of the plural “condyles” throughout the patent specification, “the importance of having the same radii on both condylar elements” to achieving the benefits of the patent, and the testimony of the inventor, the district court concluded that in a bicondylar prosthesis, where both condyles are replaced, “both condyles must meet [the geometric] requirements.” J.A. at 37. Although this is a close case, we disagree.
“It is a ‘bedrock principle’ of patent law that ‘the claims of a patent define the invention to which the patentee is entitled the right to exclude.’ ” Phillips v. AWH Corp., 415 F.3d 1303, 1312 (Fed.Cir.2005) (en banc) (quoting Innova/Pure Water, Inc. v. Safari Water Filtration Sys., Inc., 381 F.3d 1111, 1115 (Fed.Cir.2004)). The plain language of claim 15 requires only one condylar element. The femoral component must include “at least one condylar element,” which the district court correctly understood to mean “one or more.” The claim then requires that “the condylar element” meet the specified geometric limitations. See '100 patent col. 7 11.4-5, 6-7, 9. If the patentee had intended both condyles in a bicondylar prosthesis to meet these limitations, he could have drafted claim 15 to require that “both condylar elements” do so. The more natural way of drafting the claim language to achieve that result would be to require “each condylar element,” rather than “the condylar element,” to conform to the constant-radius geometry.
Wright makes several arguments in an attempt to show that, despite the plain language of the claim, such a prosthesis is not covered. First, it argues that because the purpose of the claim is “to accomplish articulation of the knee prosthesis throughout a range of flexion,” and because both condyles are involved in articulation, both condyles should be required to meet the geometric limitations. Here we are dealing with a Jepson claim. The “articulation” requirement of the preamble is a claim limitation. See Epcon Gas Sys., Inc. v. Bauer Compressors, Inc., 279 F.3d 1022, 1029 (Fed.Cir.2002). However, it is a separate limitation from the geometric requirements appearing in the body of the claim. Wright does not assert that the articulation requirement is not satisfied. In other words, Wright does not argue that a bicondylar prosthesis in which only one condyle has the required constant-radius geometry does not achieve articulation throughout the primary range of flex-ion, which the district court construed as “- 15° of hyperextension to + 75° of flexion.” J.A. at 40. A prosthesis can satisfy the articulation requirement without satisfying the geometric requirements. Thus, the [1345]*1345mere fact that both condyles are involved in articulation of the prosthesis does not require a construction of claim 15 in which both condyles must conform to the specified geometry.
Wright next argues that a bicondylar prosthesis in which only one condyle meets the geometric limitations of claim 15 will not achieve several of the purposes of the invention as a whole, recited in the specification of the '100 patent. As we have explained, such an argument is not persuasive:
The court’s task is not to limit claim language to exclude particular devices because they do not serve a perceived “purpose” of the invention.... An invention may possess a number of advantages or purposes, and there is no requirement that every claim directed to that invention be limited to encompass all of them.
E-Pass Techs., Inc. v. 3Com Corp., 343 F.3d 1364, 1370 (Fed.Cir.2003) (footnote omitted); see also Kim v. ConAgra Foods, Inc., 465 F.3d 1312, 1319 (Fed.Cir.2006) (refusing to limit claim based on “one object of the invention”); Brookhilh-Wilk 1, LLC v. Intuitive Surgical, Inc., 334 F.3d 1294, 1301 (Fed.Cir.2003). As discussed above, the invention described by claim 15 serves multiple purposes, including a simplified configuration for ease of manufacturing, increased reliability of the prosthesis, reduced stress on the knee, and an increased area of contact between the condylar element and the bearing member of the tibial component. Wright does not attempt to show that a prosthesis in which only one condylar element has the recited geometric characteristics would not achieve any of these objectives. Indeed, Wright’s own advertising for its accused product cites at least one of the same objectives, namely “providing] greater ti-bio-femoral conformity with maximum contact area.” J.A. at 2437.4
Wright also argues, however, that the specification of the '100 patent does require both condylar elements of a bicondylar prosthesis to conform to the geometric limitations of claim 15. Wright correctly points out that every disclosure of a bicondylar knee in the specification shows two condyles each meeting the geometric requirements of claim 15., However, we have repeatedly held that the fact that the specification describes only a single embodiment, standing alone, is insufficient to limit otherwise broad claim language. See Phillips, 415 F.3d at 1323; Liebel-Flarsheim Co. v. Medrad, Inc., 358 F.3d 898, 906 (Fed.Cir.2004); Teleflex, Inc. v. Ficosa N. Am. Corp., 299 F.3d 1313, 1327-28 (Fed.Cir.2002). This is not a case in which the specification makes clear that the invention requires two condyles meeting the specified geometry. Compare Ormco Corp. v. Align Tech., Inc., 498 F.3d 1307, 1313-14 (Fed.Cir.2007) (limiting claims in part based on specification’s identification of the “primary objective” of the invention and its distinction of the invention from the prior art based on that objective); SciMed Life Sys., Inc. v. Advanced Cardiovascular Sys., Inc., 242 F.3d 1337, 1342-44 (Fed.Cir.2001) (limiting claims based on language in specification repeatedly stating that “the invention” required such limitation and that the limitation was applicable to “all embodiments of the present invention”). Again, Wright acknowledges that although a natural condyle may not meet the geometric limitations of claim 15, a unicondylar prosthesis used with such a natural condyle is none[1346]*1346theless within the scope of the claim. Nothing in the specification supports a reading of claim 15 to include a unicondylar prosthesis but to exclude a bicondylar prosthesis in which only one condylar element has the required geometric characteristics.
Wright next argues that the prosecution history of the '100 patent requires its proposed construction. In making this argument, Wright relies on a letter sent by the attorney prosecuting the '100 patent to his client after an examiner interview. In this letter, the attorney explained the examiner’s position that “the claims could be worded more clearly to bring out the fact that the constant radius in the present device does extend along enough of each condylar element to cover the range of articulation specified.” J.A. at 1870 (emphasis added). Wright argues that the attorney’s use of the word “each” demonstrates that both the attorney and the examiner understood that both condylar elements in a bicondylar prosthesis must have the required geometry. Whatever meaning it might have, this letter is not part of the prosecution history, which consists of “all express representations made by or on behalf of the applicant to the examiner to induce a patent grant.” Standard Oil Co. v. Am. Cyanamid Co., 774 F.2d 448, 452 (Fed.Cir.1985). A letter between a prosecuting attorney and a patent applicant is not a representation to the patent examiner, and is therefore not part of the prosecution history. Rather, it is extrinsic evidence, which can be of limited value to claim construction, “considered in the context of the intrinsic evidence,” if it “can help educate the court regarding the field of the invention and can help the court determine what a person of ordinary skill in the art would understand claim terms to mean.” Phillips, 415 F.3d at 1319. The letter Wright relies on, reporting to the inventor on the results of an examiner interview, achieves neither of these objectives and is of no value to the construction of the disputed claim language.
Finally, Wright argues that the testimony of Dr. Mark Kester, the lead inventor of the '100 patent, compels a construction of claim 15 that requires both condylar elements in a bicondylar prosthesis to meet the geometric limitations. Dr. Kester testified that the language “at least one condylar element” was added to the claims so that they would clearly cover unicondylar prostheses. He further testified that it was his “intention that both condylar elements [in a bicondylar prosthesis] would meet the requirement[ ] of having the constant anterior-posterior radius throughout the primary range of flexion.” J.A. at 1668. The testimony of an inventor “cannot be relied on to change the meaning of the claims.” Markman v. Westview Instruments, Inc., 52 F.3d 967, 983 (Fed.Cir.1995) (en banc), aff'd, 517 U.S. 370, 116 S.Ct. 1384, 134 L.Ed.2d 577 (1996); see also Hoechst Celanese Corp. v. BP Chems. Ltd., 78 F.3d 1575, 1580 (Fed.Cir.1996) (“Markman requires us to give no deference to the testimony of the inventor about the meaning of the claims.”). In particular, we have explained that “[t]he subjective intent of the inventor when he used a particular term is of little or no probative weight in determining the scope of a claim.” Markman, 52 F.3d at 985.
Wright argues that inventor testimony is relevant in cases where the inventor does not seek to enlarge the scope of the claims to cover an accused product, but rather admits that the claims are limited to exclude that product. Under such circumstances, according to Wright, the testimony is reliable because it is against the inventor’s interest. We reject this distinction. Whether an inventor’s testimony is consistent with a broader or narrower claim scope, that testimony is still limited [1347]*1347by the fact that an inventor understands the invention but may not understand the claims, which are typically drafted by the attorney prosecuting the patent application. As we have explained, “it is not unusual for there to be a significant difference between what an inventor thinks his patented invention is and what the ultimate scope of the claims is after allowance by the PTO.” Id. Moreover, Wright’s asserted approach, to rely on inventor testimony when it is contrary to interest, is unworkable. It would require a case by case determination as to whether an inventor is testifying against his or her interest. The inventor might testify to a broad claim scope in order to increase the likelihood of a finding of infringement. The inventor also might testify to a narrower claim scope to avoid a challenge to the validity of the patent. We hold that inventor testimony as to the inventor’s subjective intent is irrelevant to the issue of claim construction.5
As explained above, the plain language of claim 15 makes clear that, whether a prosthesis is unicondylar or bicondylar, only one condylar element meeting the geometric limitations of the claim is required to bring that prosthesis within the scope of the claim. Wright offers no persuasive evidence that the claim should be interpreted other than by its plain language. Because we conclude that the claim requires only that one condylar element in a bicondylar prosthesis satisfy the claim limitations, we vacate the judgment and remand.
As a separate and independent ground, Wright asserts that the judgment may nonetheless be affirmed on the basis that Howmedica’s infringement suit is barred by the release provision in the New Jersey agreement. Wright argues that the district court incorrectly granted summary judgment in favor of Howmedica on the release defense because New Jersey law does not support a construction of the contract language here to exclude the present action from the release. We review the district court’s summary judgment with regard to the release provision de novo. See Symantec Corp. v. Computer Assocs. Int’l, Inc., 522 F.3d 1279, 1287 (Fed.Cir.2008). Construction of a contract is an issue of law that we review without deference. See Intel Corp. v. VIA Techs., Inc., 319 F.3d 1357, 1361 (Fed.Cir.2003). We apply state contract law in interpreting a settlement agreement. See Augustine Med., Inc. v. Progressive Dynamics, Inc., 194 F.3d 1367, 1370 (Fed.Cir.1999).
The district court concluded, “[a]fter reviewing all the facts and circumstances surrounding the creation of the Companion Settlement Agreements, as well as the actual language used, ... the instant action is not barred by the release provision.” J.A. at 63. We agree with the result reached by the district court, although we rest our decision on slightly different grounds.
The district court found that “it was the intent of the parties that the settlement release only cover the four lawsuits that were the subject of the Companion Settlement Agreements.” Id. However, on its face, the release provision of the New Jersey agreement applies to “any and all manner of claims ... that Howmedica ... has, ha[s] had, or may have against Wright Medical ... including, but not limited to, any and all claims and counterclaims that [1348]*1348were or could have been asserted” in the New Jersey lawsuits. Id. at 302, ¶ 27. While releases generally are construed not to release future claims, see Smith v. Smith, 72 N.J. 350, 371 A.2d 1, 6 (1977); Bilotti v. Accurate Forming Corp., 39 N.J. 184, 188 A.2d 24, 35-36 (1963), this language would appear to release the claims in question since they existed at the time of the settlement agreement. See Augustine, 194 F.3d at 1372 (construing similar language broadly to “release[ ] all possible future claims”); Three Rivers Motors Co. v. Ford Motor Co., 522 F.2d 885, 895 (3d Cir.1975) (finding that similar language left “nothing open and unsettled between the parties”).
New Jersey law does allow a court interpreting a contract provision to consider extrinsic evidence. See Halper v. Halper, 164 F.3d 830, 840-41 (3d Cir.1999); Atl. N. Airlines v. Schwimmer, 12 N.J. 293, 96 A.2d 652, 656 (1953). “Such evidence is adducible only for the purpose of interpreting the writing — not for the purpose of modifying or enlarging or curtailing its terms.” Atl. N. Airlines, 96 A.2d at 656. We agree with the district court that the intent of the parties was to have the same release language in both the New Jersey and the Massachusetts contracts. We need not reach the question of whether, under New Jersey law, this result can be achieved through “interpretation” of the contract because we conclude that reformation of the contract is appropriate to conform the language of the New Jersey agreement to match the language of the Massachusetts agreement and thus reflect the actual intention of the parties.6
The undisputed facts show that reformation would be available even if we were to reject the contract interpretation analysis by the district court as being in reality a modification of the contract. When a written contract “fails to express the agreement because of a mistake of both parties as to the contents or effect of the writing, the court may at the request of a party reform the writing to express the agreement.” Restatement (Second) of Contracts § 155 (1981). The purpose of reformation, therefore, is to correct a mistake that occurred in reducing the parties’ actual, negotiated agreement to writing. “The mere fact that a mistaken party could have avoided the mistake by the exercise of reasonable care does not preclude ... reformation,” unless that party’s “fault amounts to a failure to act in good faith and in accordance with reasonable [1349]*1349standards of fair dealing.” Id. § 157 & cmt. a. Reformation of a contract requires clear and convincing evidence of “a mistake common to both parties” that causes the written instrument not to reflect the real agreement between the parties. Brodzinsky, 182 A.2d at 153. Moreover, “the mistake must be a material one; it must be one in the absence of which the party who made it would not have entered into the compromise.” 28 Richard A. Lord, Williston on Contracts § 70:219 (4th ed.1990).
The two agreements, drafted by Wright’s counsel, originally had identical language. The testimony of both parties’ representatives made clear that the parties understood that the two agreements would have the same effect. Indeed, Wright’s President and Chief Executive Officer, Patton, testified that at the time of the settlement he did not realize that there was a difference between the language of the New Jersey and Massachusetts release provisions. Likewise, Zarnowski, Howmedica’s Senior Director of Intellectual Property, testified that when he reviewed the final drafts of the two agreements, he “did not discover the differences between the two release clauses.” J.A. at 620. In the district court proceedings, Wright conceded that, presumably, “if [Howmedica had] asked for the change in the [New Jersey] agreement” that had been made in the Massachusetts agreement, “the change would have been made.” J.A. at 767-68. Wright has identified no contrary evidence showing that the parties intended the two provisions to have different language or to be of different scope, or that there was any plausible reason as to why the parties would wish the New Jersey release provision to have a broader scope than the Massachusetts release provision. The testimony of Patton and Zarnowski demonstrates that the parties were mutually mistaken in believing there to be no difference between the two release provisions in the final drafts of the agreements. It is significant in this connection that the parties in the integration clause of the agreements indicated that each agreement would be construed to be consistent with the other — i.e., that the agreements were not designed to conflict, as would be the case if the two agreements had different release provisions. It is also significant that Wright did not even assert its release defense until Howmedica’s action had been pending for over two years, suggesting a belated discovery of a drafting error. See Journeymen Barbers v. Pollino, 22 N.J. 389, 126 A.2d 194, 197 (1956) (noting that “practical interpretation” evidenced by parties’ post-performance conduct was “entitled to great weight”); Wheatly v. Sook Suh, 217 N.J.Super. 233, 525 A.2d 340, 345 (App.Div.1987) (relying on “the conduct of defendants after the execution of the agreement”). This mutual mistake was material. Accordingly, based on the parties’ mutual and material mistake, the release provision of the New Jersey agreement should be reformed to have the same language as the Massachusetts release provision.
The question remains, however, what the meaning of that language is. Wright contends that even under the narrower language of the Massachusetts agreement, the current suit is barred. As reformed, the release provision applies to claims “that were or could have been asserted by Howmedica” in the earlier litigation. J.A. at 382. The meaning of this language, by itself, is ambiguous, and could be read to cover any infringement claim where the patent had issued and the accused product existed before the date of the agreement, or any infringement claim for which the patentee had completed a reasonable presuit investigation to allow a good faith filing of the claim, or only a claim that had matured to the point where both parties recognized the existence of a controversy [1350]*1350that would be a predicate to filing an infringement suit.
Here the district court properly relied on the intent of the parties to interpret the ambiguous language. When a contract provision is ambiguous, New Jersey law permits reference to extrinsic evidence to determine the intent of the parties. See Halper, 164 F.3d at 841; Krosnowski v. Krosnowski, 22 N.J. 376, 126 A.2d 182, 187-88 (1956); Atl. N. Airlines, 96 A.2d at 656. “When it becomes clear that a ‘certain factual result’ was within the contemplation of the parties, ‘interpretation should be affected by reasonable and necessary implications, so that the legal effect then given to the instrument will be such as to attain the intended factual result.’ ” Krosnowski, 126 A.2d at 188 (citation omitted).
The evidence makes clear that the parties did not intend the release provision to cover the current infringement action. As the district court noted, the primary purpose of the agreements was to resolve the pending litigation. Cross-licenses of the four asserted patents were made a part of the agreement. Although a license for use of another patent, the Averill patent, was included, Howmedica did not grant Wright any license under the '100 patent. Given the structure of the agreements it is on its face unlikely that the parties would exclude the '100 patent from the cross-licensing provisions but effectively grant rights in the patent to Wright with regard to existing products by virtue of the release provision.
The testimony of both parties confirms this reading of the agreements. Lipes (the President of Howmedica) and Zarnowski (Howmedica’s Senior Director of Intellectual Property) both testified that Howmedica refused to agree during the settlement negotiations to grant Wright any rights with respect to the '100 patent or the ADVANCE® knee product. Zarnowski explained that both he and Lipes “had an understanding that there was not a dispute” with regard to the '100 patent, and that therefore “it was not part of the release.” J.A. at 621. Patton (Wright’s President and Chief Executive Officer) admitted that he did not believe the release provision applied to patents owned by the parties that were unrelated to the matters then in dispute, or to claims unrelated to those matters. Although Wright argues, and Patton’s testimony confirms, that Wright was seeking what it called a “global settlement,” Patton acknowledged that his understanding of the release provision was that it covered only current disputes between the parties. Although he knew of and had discussed with Lipes rumors that Howmedica might have a claim against the ADVANCE® knee product, Patton stated that his view was that this kind of “unsubstantiated allegation” did not constitute an actual controversy covered by the release provision. J.A. at 918.
Thus, we conclude that both parties intended that the reformed release language be limited to matters actually in dispute at the time. Accordingly, the district court correctly granted summary judgment in favor of Howmedica on Wright’s affirmative defense based on the release. We therefore reject Wright’s assertion that the defense is an alternative ground on which to affirm the district court’s judgment.
CONCLUSION
For the reasons discussed above, we vacate the judgment of nonmfringement, and remand to the district court for further proceedings consistent with this opinion.
VACATED AND REMANDED
[1351]*1351COSTS
No costs.
Circuit Judge PROST dissents.