WIENER, Circuit Judge:
The plaintiff-appellants in this maritime action were the owners and operators of the M/V KAY ECKSTEIN (the “KAY”) and their insurers (collectively, “Plaintiffs”). The KAY was a triple screw, steel hulled push boat which was originally constructed in 1973. In 1999, the KAY’s three Caterpillar 3606 engines, which had been installed in 1992, were overhauled and other extensive renovations were performed by defendant-appellant Quality Shipyards, Inc. (“Quality”) and defendant Louisiana Machinery Company, Inc. (collectively, “Defendants”). After the work was completed, the KAY successfully underwent dock and sea trials and was returned to Marquette, which operated the vessel for five weeks without incident. Late in May 1999, however, a catastrophic engine-room fire led to the KAY’s total loss.
Plaintiffs sued Defendants for maritime negligence and breach of express and implied warranties of workmanlike service. Quality counterclaimed seeking attorneys’ fees and court costs. After a bench trial, the district court concluded that (1) Marquette had not carried its burden of proof with regard to negligence and causation, but (2) Quality’s counterclaim was without merit, notwithstanding the repair agreement’s indemnity provision. Both sides appealed. Although we see no reason to disturb the district court’s disposition of the negligence and breach of contract claims, we conclude that the trial court erred in interpreting the indemnity provision here at issue and in denying Quality’s counterclaim.
I. Facts and Proceedings
The KAY underwent extensive maintenance and repair work in 1998-99, including the overhauling of its three Caterpillar engines; designing and fabricating kort nozzles and I struts around its screws; removing, modifying, and reinstalling the screws, rudders, and shaft lines; and testing and delivering the vessel. The engine overhauls included installing new fuel and oil filters and resealing and rebushing the oil pump. The contract price for the work was $870,000.
None of the KAY’s crewmembers witnessed the start of the fire, although one
crewman had been in the engine room thirty minutes prior to the fire’s estimated start time. The KAY lost engine propulsion approximately thirty minutes after the fire was first noticed, and sank shortly thereafter. After the KAY was raised from the river approximately 10 days after it had sunk, Plaintiffs discovered that the check valve fittings on the center main engine were loose.
At trial, Plaintiffs’ primary contention was that Defendants had improperly torqued (tightened) the check valve fittings on the KAY’s center main engine. According to Plaintiffs’ theory, those under-tightened fittings had been gradually loosened by engine vibration, eventually allowing a fuel spray to develop. Plaintiffs advanced several potential ignition sources, the most probable of which — according to Plaintiffs — was the exhaust pipe of the generator’s diesel engine. When the trial ended, the district court concluded that Plaintiffs had not carried their burden of proof with regard to either fault or causation, and held for Defendants. Plaintiffs argue on appeal that the district court erred as a matter of law by holding Plaintiffs to an improperly high burden of proof.
Quality counterclaimed for the attorneys’ fees and costs it incurred in defending the suit, basing its claims on the repair agreement’s indemnification provision, which by its terms applies to such expenses:
Each party agrees to defend, indemnify and hold harmless the other party’s in-demnitees free and harmless from and against any and all suits, claims, or liabilities (including, without limitation, the cost of defending any suit and reasonable attorney’s fees).
When it first considered the indemnity provision, the district court granted Quality’s motion to exclude parol evidence on the question of the parties’ intent, holding that the terms of the provision were unambiguous. The court denied Quality’s motion for summary judgment on this claim, however, indicating that the indemnification provision would not be enforceable if the Defendants had acted with gross negligence. The district court subsequently ruled that although Quality was not negligent in this matter, the indemnity provision had to be interpreted in conjunction with the other provisions of the repair agreement. As the agreement required each party to obtain specified insurance policies, concluded the district court, the proceeds of those policies were intended to be the “primary payer” of the subject damages, ahead of the contract’s indemnity obligations: “There is no logical way to reconcile the indemnity provisions and the mandatory insurance provisions ... other than to find that the parties intended that the insurance coverages be exhausted pri- or to the indemnity obligation being triggered.”
In reaching this conclusion, the district court relied on
Ogea v. Loffland Brothers
Company
and
Tullier v. Halliburton Geophysical Services,
Inc.,
cases in which each party to an indemnity agreement was required to name the other as an “additional insured” under the mandated insurance policies. Quality argues that because there was no requirement in the instant contract to name the party opposite as an
additional insured, Marquette was not entitled to any benefit from the insurance policies at issue. As such, insists Quality, it should be reimbursed by Marquette under the terms of the indemnification provision without regard to those insurance policies. The district court took a “broader view,” finding that the “existence of mandatory reciprocal insurance obligations” was determinative, despite the fact that, unlike the situations in
Ogea
and
Tullier,
the instant obligations did not require that the other party be named as an additional insured.
Quality thus maintains that the district court’s ruling constitutes an unwarranted extension of our precedent.
II. Analysis
A.
Standard of Review
We review the district court’s conclusions of law — including its contractual in
terpretations
— de
novo.
We review findings of fact for clear error.
B.
Burden of Proof
Plaintiffs must show negligence and causation by a preponderance of the evidence.
Marquette correctly notes that in a fire case, these elements frequently must be established by circumstantial evidence because of the fire’s destruction of the physical evidence.
Even so, the evidence available must be
sufficient
to find
both
negligence and causation.
We address each in turn.
I.
Negligence
The district court correctly noted that, to assess whether Defendants were negligent, the court first had to “determine whether it is more probable than not that the couplings were loose at the time the vessel left the shipyard or whether vibration or heat from the fire likely loosened the fittings.”
This task was complicated by the fact that the KAY had been under water, exposed to river currents and elements, for some 10 days before it was raised and examined. Furthermore, when Plaintiffs’ representative examined the KAY and reported that the check valve fittings at issue were loose, he neglected to mark the fittings in any way that would record for posterity just
how
loose they were. Instead, Plaintiffs’ representative relied on “his degree of ‘gentle shaking’ and his ability to rotate the fittings” to estimate their condition at the time of the fire,
but this “measurement” was performed after the fittings had been subjected to (1) the extreme heat of the fire and (2) the currents of the river.
Because of these factors, the district court relied on circumstantial evidence to extrapolate the status of the fittings as of the time that the KAY had left the shipyard. The court noted that during the KAY’s five weeks of operation prior to the fire, there had been no evidence of a loose check valve fitting; that is, no crew member ever saw fuel seeping or spraying from the connection.
Therefore, reasoned the
district court, one important question is “if the fittings were improperly tightened before the vessel left the shipyard, would the fittings have leaked at some point after the ship left the shipyard and prior to spraying fuel on the date of the fire?”
On this question, there was trial testimony from more than one source that if the fittings were loose because they had been improperly torqued, a leak likely would have developed as soon as the engine was cranked and pressure built up.
Further complicating Plaintiffs’ negligence theory is the fact that it is not clear
who
reinstalled the fuel transfer line after the engine mountings were drilled. First, there was no notation on any written record indicating that Defendants’ employees had performed that task.
Neither did the check valve fittings themselves need to be loosened for the fuel pump to be reinstalled. Finally, record evidence indicates that Marquette personnel — including an engineer — were at the site when the work was completed. As those employees had previously performed some work on the engine, it is at least possible that Plaintiffs’ own employees participated in the reinstal-lation of the pump. Given the lack of evidence suggesting that the check valve fittings were loose when the KAY left the shipyard, and the lack of evidence that Defendants’ employees alone were responsible for improperly torquing the fittings if they were in fact loose, the district court concluded that “Plaintiffs have not proved by a preponderance of the evidence that there was any contractual breach or maritime negligence by either defendant.”
2.
Causation
The district court was equally unimpressed by Plaintiffs’ causation theory. Experts on both sides testified that, over time, engine vibration could further loosen an improperly torqued fitting, eventually leading to a spray of fuel; and that it was hard to know how long it might take for such a spray to develop. Defendants, however, produced experts with significant experience working with the flared valve fittings here at issue (experience the Plaintiffs’ experts did not have), who ventured that it would be “highly improbable” for loose fittings to move from no leak to a full-blown spray within the 30-minute period involved in this case.
Defendants created a model of the center main engine to demonstrate that even a check valve fitting improperly torqued to the degree
alleged
by Plaintiffs would not produce a fuel spray that could reach what Plaintiffs identified as the most likely source of ignition — the diesel generator exhaust pipe.
Defendants also produced a fire expert who stated that the absence of a specific “flash pattern” indicated that the fire did not start in the area alleged by Plaintiffs.
In other words, even if the check valve fittings were loose when the
KAY left the shipyard, it is unclear that those fittings
could
have caused the fire under these circumstances, much less that in fact they did so. The district court found for Defendants on this issue, concluding that “Plaintiffs have also not met their burden of proof with respect to causation.”
3.
Plaintiffs’ contentions on appeal
On appeal, Plaintiffs argue that the circumstantial evidence they produced was sufficient for a finding of liability by a preponderance of the evidence, given how that standard has been interpreted in the applicable case law. The district court, argue Plaintiffs, held them to an inappropriately high burden of proof on negligence and causation. For support, Plaintiffs cite cases discussing the burden of proof in fire cases, which stand for the propositions that (1) circumstantial evidence may support a finding of negligence and causation,
(2) a plaintiffs proof need not exclude or eliminate every other possible cause of the fire,
and (3) a plaintiff need not establish the method or point of ignition, but only sufficient circumstances implicating the defendant.
Plaintiffs’ arguments on this issue ultimately fail, as all the cases that they cite contain circumstances allowing for strong inferences of negligence and causation— circumstances not present in the instant case. In
Boudreaux v. American Insurance
Company,
for example, the restaurant that burned down was under the exclusive control of the defendants on the evening of the fire, which started after hours. Similarly, in
Hanover Insurance Company v. Jacobson-Young, Inc.,
a fire broke out because Jacobson-Young employees had improperly stored flammable materials in an area where
only
Jacobson-Young employees were allowed. The other cases cited by Marquette are similarly distinguishable.
Additionally, all the cases ultimately detail the same burden of proof, which the plaintiff meets when “the inferences from the testimony are such as to persuade that the occurrence of an essential fact was more likely or probable than its non-occurrence.”
In the cases cited by Plaintiffs, the fact patterns were such that the circumstantial evidence was sufficient to find liability. Here, in contrast, (1) there was credible expert testimony on both sides, (2) the KAY had been out of Defendants’ control for more than a month before the fire, and (3) the destruction caused by the fire made it difficult to discern the cause. In combination, these facts made it difficult for Plaintiffs to prove their theory of the accident, even by a preponderance. And, in its capacity as the finder of fact, the district court concluded that Plaintiffs had failed to carry this burden, labeling their theory as “improbable,” and “possible, but unlikely.”
Plaintiffs make two additional arguments about the district court’s methodology, both revolving around the claim that the court put them to an improperly high burden of proof. Plaintiffs argue, for example, that the burden shifted to Defendants to propose an equally probable cause of the fire after Plaintiffs met their burden by demonstrating negligence. This argument, of course, presumes that Plaintiffs actually met their burden in the first place, and is undermined by the district court’s facially logical — and apparently proper— conclusion that they had not done so. Similarly, Plaintiffs argue that the district court improperly “impos[ed] ... a requirement to prove the exact mode and point of ignition.” The district court did note that it “rejected] the notion that such a [fuel] spray permeated the elbow joint of the diesel generator exhaust in an amount sufficient to ignite.”
This statement, however, is merely part of the district court’s discussion of the strengths and weaknesses of Plaintiffs’ case. And, that discussion of the ignition point matters little in the overall context of this case, as the district court also found that the
other
elements of Plaintiffs’ fire theory were improbable and that there was no contractual breach or maritime negligence by Defendants in any event.
In sum, we perceive that the district court did weigh the conflicting evidence— all of it credible — and concluded that Plaintiffs had not proved their theory of the accident by a preponderance of the evidence. That there are many fire cases in which the plaintiff was able to bear his burden with purely circumstantial evidence does not automatically make the district court’s approach — or its conclusion — erroneous.
C.
Indemnification
Quality’s counterclaim, as noted, is based on the repair agreement’s indemnification clause, which binds the signatories, Quality and Marquette. That clause provides, in relevant part:
Each party agrees to defend, indemnify and hold harmless the other party’s In-demnitees free and harmless from and against any and all suits, claims, or liabilities
(including, without limitation, the cost of defending any suit and reasonable attorney’s fees)
for loss or damage to property owned, leased or operated by the indemnitor,
regardless of cause, including the negligence or other legal fault of any of each party’s Indem-nitees,
Quality argues that “property owned, leased or operated” by Marquette includes the KAY itself. Therefore, according to Quality, because Marquette wrongfully sued for the loss of its “property,” Marquette should be responsible for the expenses Quality incurred in defending the claim.
As we have noted, the district court’s decision on this matter turned on the interplay between the indemnification clause and the reciprocal insurance obligations required by the repair agreement. Specifically, Quality was required to purchase:
at its own expense for its own employees, properties and operations, the following policies of insurance:
(A) By Shipyard—
(1) Worker’s Compensation ... and employer’s liability insurance and/or appropriate maritime employers coverage
(2) Comprehensive Public Liability and Ship Repairers’ Liability Insurance ... including broad form contractual liability coverage ...;
(3) Automobile liability insurance ...; and
(4) Full form physical damage insurance on all property (including floating equipment and vessels) owned, chartered, operated, or otherwise used by the Shipyard.
The repair agreement contains similar insurance requirements for Marquette, with additional required coverages.
The district court looked to a series of cases in which we held (or affirmed) that, because of reciprocal insurance requirements, contractual indemnity provisions did not apply until the limits of those insurance policies had been reached. In other words, the insurance policies were the “primary payers” and should be exhausted before any indemnity obligations attached. In those
cases
— Ogea
v. Loffland Brothers Company,
Tullier v. Halliburton Geophysical Services, Inc.,
and
In Re Diamond Services
— the contractual insurance obligations included the express requirement that each party name the other as an additional insured under the applicable policies. In
Tullier,
we noted that this factor had been “controlling” in
Ogea,
and deemed it so again in
Tullier.
In
Diamond Services,
the district court explained that although there was no additional-insured requirement for the Comprehensive General Liability policy, which contained the contractual liability coverage, there
was
such a requirement for the P&I policy, which provided primary coverage. The district court in
Diamond Services
explained that the “essential fact is that the liability insurance provided ... under Diamond’s P&I policy is primary and, therefore, before CMC is required to indemnify Diamond, the limits of the P&I policy must be exhausted.”
In the instant case, there are no contractual provisions requiring “additional insured” coverage or, as in
Diamond Services,
dictating that the contractually-required insurance policies provide primary coverage. The district court nevertheless looked past this fact, taking a “broader view” and examining the underlying reasoning of
Ogea
and
Tullier.
As the district court noted, in both of those cases we explained that it is necessary to read all contractual provisions “in conjunction with each other in order to properly interpret the meaning of the contract.”
This mandate, combined with the following reasoning from
Diamond Services,
led the district court to conclude here that the presence of “additional insured” coverage is not a critical factor:
There is no reason for an indemnitor to require an indemnitee to procure insurance if the indemnitor did not intend to limit its indemnification obligations to the excess of the required insurance coverage .... To read the indemnity and insurance requirements any other way produces an incoherent result, e.g., why would CMC require Diamond to obtain certain insurance policies if CMC is required to indemnify Diamond for any claims covered under those policies.
The district court explained that, “[ajfter considering the repair agreement as a whole,” it came to the same conclusion as did the
Diamond Services
court: “There is no logical way to reconcile the indemnity provisions and the mandatory insurance provisions ... other than to find that the parties intended that the insurance coverages be exhausted prior to the indemnity obligation being triggered.”
We disagree with the district court’s conclusion that there is “no logical way” to reconcile the indemnity and insurance provisions without finding that the parties intended that the insurance limits be exhausted prior to attachment of the indemnity obligations. This line of thinking is appropriate in cases like
Ogea, Tullier,
and
Diamond Services,
where the contracts contained “additional insured” requirements or dictated the primacy of insurance coverage over indemnification obligations, or both. That made it illogical to read the underlying contracts as providing for anything
other
than an indemnity obligation that does not become operable until and unless the insurance proceeds are exhausted. In the absence of similar contractual language, however, this reasoning is inap-posite. In the instant case, the insurance requirement appears to be designed to provide a solvent, deep pocket for any indemnity obligations that may eventuate between the parties and to cover any third-party claims that might arise. It simply is not true that there is only one way to integrate the repair agreement’s indemnity and insurance obligations. In this absence of language supporting the district court’s interpretation of those provisions, we cannot accept it.
Marquette also makes a number of alternative arguments on the indemnity clause. Although all these arguments are based, at least in part, on the assumption that the indemnity clause is ambiguous, and have therefore been either directly or inferentially addressed by the district
court’s decision on summary judgment,
we touch on each briefly.
First, Marquette argues that an ambiguity exists in the “other parties’ indemni-tees” language of the indemnity clause. Specifically, Marquette notes that “the vessel ... its registered owner, and each of their respective underwriters” are included in the definitions of both “owner indemnitees” and “shipyard indemnitees.” Marquette asserts that this makes the provision “confusing and contradictory as to who is indemnifying whom.” Although, certainly, an indemnity claim by the vessel’s owners against themselves would be contradictory and confusing, we find no ambiguity in the contract’s requirement that Marquette indemnify Quality. “Shipyard, its parent, subsidiary, and affiliated companies” are the first of the listed “Shipyard indemnitees.” That fact, in combination with the phrase “[e]ach party agrees to defend, indemnify, and hold harmless the other party’s indemnitees,” makes clear that Marquette agreed to indemnify Quality and hold it harmless.
Marquette also argues that the indemnity clause amounts to an unenforceable exculpatory clause. Again, we disagree. First, the clause clearly indicates that the indemnification obligation will attach regardless of the negligence of any of the indemnitees: “Each party agrees to defend, indemnify and hold harmless the other party’s indemnitees ... regardless of cause, including the negligence or other legal fault of each party’s indemnitees.” Second, this particular indemnity clause does not absolve Quality of its warranty duties under the contract as Marquette alleges.
If Quality had been found to be in breach of those warranties, perhaps our application of the indemnity clause would be different. As Quality apparently met its duties under the contract, however, it is entitled to indemnification from Marquette for the costs and expenses caused by the latter’s suit.
Marquette contends in addition that the phrase “property owned, leased, o.r operated” does not encompass the vessel itself. In an effort to support this proposition, Marquette notes that the term “vessel” is used 19 times in the repair agreement, but the term “property” is used only six times, and argues this is an indicator that different meanings are ascribed to the two terms. Although the limited use of the term “property” may suggest a limited meaning, the plain fact is that, as between the terms “property” and “vessel,” the former is the broader of the two. “Vessel” is a lesser included
type
of “property,” but it is property nonetheless. If Marquette had wanted to exclude the vessel from the ambit of “property,” it could have insisted on language such as “property other than the vessel” or even “other property” in the indemnity clause.
No such language is present, however, and we will not infer limiting language absent any indication of party intent.
Finally, Marquette asserts that the indemnity clause should be construed
against its drafter — Quality—again, because of the clause’s alleged ambiguity. Beyond our conclusion, shared by the district court, that the language is not ambiguous (certainly not in the context of the instant case), we note further that changes — initialed by both parties — were made to the indemnification provisions found in paragraph 8 of the repair agreement.
This shows that Marquette read and considered the language of the indemnification clause — again, language that we do not find ambiguous. Under these facts, we decline to impose a strained construction of the language on Quality for having drafted the initial version of the agreement.
III. Conclusion
The district court did not require Plaintiffs to meet an incorrectly difficult burden of proof on the issues of negligence and causation; rather, Plaintiffs simply failed to carry the proper burden. On the issue of indemnification, however, the district court’s expansive application of the
Ogea /Tullier
reasoning to the instant situation is unwarranted. The repair agreement in this case — unlike those in
Ogea
and
Tullier
— did not require that any party opposite be named as an additional insured; neither did it dictate that the required insurance would provide primary coverage before indemnification. Absent explicit language entitling Marquette to benefit from the proceeds of those insurance policies, we see no justification for reading such provisions into the agreement. ' We therefore affirm that portion of the district court’s August 6, 2002 Order finding that Plaintiffs had not carried their burden of proof with regard to negligence and causation, but reverse that portion of the Order finding Quality’s counterclaim without merit by virtue of the interplay between the repair agreement’s indemnification clause and insurance obligations. The decision of the district court is therefore
AFFIRMED in part, REVERSED in part, and REMANDED with instructions.