Opinion for the Court filed by Circuit Judge SILBERMAN.
Dissenting Opinion filed by Circuit Judge EDWARDS.
SILBERMAN, Circuit Judge:This is an unusual case in which both the appellant and the government present us with unreasonable interpretations of a statute we think ambiguous. We therefore direct the district court to remand to the agency for reconsideration.
I.
In 1984, Congress enacted the Drug Price Competition and Patent Term Restoration Act, Pub.L. No. 98-417, 98 Stat. 1585 (1984) (the “Hatch-Waxman Amendments”), amending the Federal Food, Drug, and Cosmetics Act, 21 U.S.C. §§ 301-392. The statute created a new system for protecting both the interests of drug manufacturers who produce new drugs and the interests of generic drug manufacturers and their consumers. Facing the classic question of the appropriate trade-off between greater incentives for the invention of new products and greater affordability of those products, Congress struck a balance between expediting generic drug applications and protecting the interests of the original drug manufacturers. See H.R. Rep. No. 98-857 (Pt. 1), 98th Cong., 2d Sess. 14, 15, reprinted in 1984 U.S.Code Cong. & Admin.News 2647, 2648. It reserved a longer period of market exclusivity for newly developed drugs than for drugs for which the approval process had already been completed in a prior application. The statute gives the original drug producer a specified period of market exclusivity depending primarily on the pharmaceutical novelty of a drug.
This case requires us to determine the scope of the exclusivity provisions, in particular the meaning of the phrase “active ingredient (including any ester or salt of the active ingredient).” 21 U.S.C. § 355(j)(4)(D)(i) and (v). The full statutory language is set out in the margin, but the key parts of the two relevant subsections of the statute state that a two year exclu*986sivity period attaches to a drug whose new drug application (“NDA”) “includes an active ingredient (including any ester or salt of the active ingredient) that has been approved in another [NDA],” 21 U.S.C. § 355(j)(4)(D)(v) and a ten year exclusivity period attaches to a drug whose NDA has “no active ingredient (including any ester or salt of the active ingredient) of which has been approved in any other [NDA],” 21 U.S.C. § SSSQ^XDXi).1 Congress thereby sought to encourage innovation in the drug industry, by rewarding a pioneer drug with a ten year exclusivity, while protecting consumers from unduly high prices by refusing to give a long period of market exclusivity to drugs which required no new research effort.
In 1978 the Food and Drug Administration (“FDA”) approved for marketing Abbott Laboratories’ (“Abbott”) new drug application for Depakene, an anticonvulsant drug prescribed for control of epileptic and other seizures. The chemical ingredient of Depakene that performs the drug’s therapeutic function is valproic acid. Valproic acid is both an “active ingredient” (the substance prior to introduction into the human body) and an “active moiety” (the substance that creates the actual therapeutic effect within the body). In 1982, the FDA approved Abbott’s new drug application (“NDA”) for another drug, Depakote, also prescribed for control of seizures. In its Depakote NDA, Abbott relied exclusively on the safety and efficacy findings established in the course of the prior Depakene application, because the “active moiety” of Depakote was the very same valproic acid. Depakote’s active ingredient, we are told, is divalproex sodium which is converted into the valproic acid within the human body. Apparently divalproex sodium presents fewer gastrointestinal side effects than the valproic acid. Although the issue was disputed, the agency found divalproex sodium to be a salt of valproic acid.2 In simple terms, a salt is a chemical compound created when the “parent” substance reacts with another chemical. A salt is “formed when the hydrogen of an acid is replaced by a metal or its equivalent.” Glaxo Operations UK Ltd. v. Quigg, 894 F.2d 392, 393 n. 3 (Fed.Cir.1990), citing The Condensed Chemical Dictionary 907 (G. Hawley rev. 10th ed. 1981). FDA determined that De-pakote could only be granted a two year period of market exclusivity pursuant to the statute because it was a salt of the active ingredient of the prior-approved De-pakene. On August 29, 1986 Abbott sub*987mitted a Citizen Petition to the FDA pursuant to 21 C.F.R. § 10.30 requesting a ten year exclusivity. The FDA denied Abbott’s petition, Abbott Labs, Docket No. 86P-0367 (FDA February 11, 1988), and Abbott filed a suit in the district court. It appeals from a judgment below which affirmed the agency decision. Abbott Labs. v. Young, 691 F.Supp. 462 (D.D.C.1988).
II.
Pursuant to the Supreme Court’s guidance in Chevron, we must first determine whether Congress manifested an “unambiguously expressed intent” that resolves this dispute over the statute’s meaning. Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 842-43, 104 S.Ct. 2778, 2781-82, 81 L.Ed.2d 694 (1984). Of course, the language of the statute itself is always the best indication of congressional intent. Abbott argues, with the support of the Federal Circuit, that the “plain meaning” of the language supports its interpretation. Both Abbott and the Federal Circuit, see Glaxo, 894 F.2d at 393-94, focus only on the phrase “active ingredient,” claiming it has a well understood meaning. The district court, on the other hand, thought that the usage of the reference words “which,” “of which,” and “that” in sections (i) and (v) created an overlap and therefore an irreconcilable conflict between the sections because the same active ingredient was ostensibly covered by both provisions, resulting in different periods of exclusivity mandated for the same substance. See 691 F.Supp. at 472. The government disavows the district judge’s perceived irreconcilable conflict.3 Instead, the government reads the parenthetical phrase (“including any ester or salt of the active ingredient”) to permit an interpretation of “active ingredient” that includes even more than salt or ester derivatives. According to the government, that phrase can be interpreted to mean that Congress was using the term active ingredient loosely, possibly as a virtual synonym for active moiety.
Indeed, it was at least suggested in an agency letter, subsequent to the decision in Abbott’s case, that the phrase “active ingredient” itself, even without the parenthetical, could be interpreted to include active moiety notwithstanding that the FDA construes that term narrowly in another section of the act, 335(j)(4)(D). See McNeil Pharmaceutical 6, Docket No. 87P-0339 (FDA July 26,1989). The agency appears to contend that the different purpose of that other section 335(j)(4)(D) (providing for expedited approval of generic drugs so agency caution is indicated) justifies a different interpretation of the same phrase. We note that it is not impermissible under Chevron for an agency to interpret an imprecise term differently in two separate sections of a statute which have different purposes. See National Ass’n of Casualty and Surety Agents v. Board of Governors of the Fed. Reserve Sys., 856 F.2d 282, 287 (D.C.Cir.1988) (upholding different agency interpretations of same phrase when based on reasonable explanation), cert. denied, 490 U.S. 1090, 109 S.Ct. 2430, 104 L.Ed.2d 987 (1989); Comite Pro Rescate v. Sewer Auth., 888 F.2d 180, 187 (1st Cir.1989) (same), cert. denied, _ U.S. _, 110 S.Ct. 1476, 108 L.Ed.2d 613 (1990). But we cannot consider whether active ingredient is such a term because the agency did not in its decision on Abbott’s application employ this theory.
Putting aside for a moment the relative merits of the various constructions offered, we first conclude the language is ambiguous as it relates to the issue before us. The parenthetical phrase (“including any ester or salt of the active ingredient”) can refer to either the active ingredient of the original approved drug or to the active ingredient in the new drug, depending on how “the” in the parenthetical and the words surrounding the parenthetical — "no active ingredient ... of which has been approved” — is interpreted. In other words, the definition of an active ingredient as including both the active ingredient and an ester or salt of the active ingredient can *988refer both to the active ingredient in the earlier and the- later drug application, which would be the proper reading if Congress had in mind, as it seems to have had, that an active ingredient was to be regarded for purpose of this portion of the statute as equivalent to an ester or salt of itself.
In this section of the Act, Congress was concerned with defining the chemical entity on which a subsection (j) (generic drug) application is based.4 When Congress used the definition “an active ingredient (or the salt or ester of the active ingredient),” § 355(j)(4)(D)(v) (emphasis added), that definition arguably referred to all the subsection (b) applications in question. The FDA’s authorization to approve a generic drug (subsection (j) application) depends on its relationship to all original drugs (subsection (b) applications). Therefore, Congress may have intended for its definition —“an active ingredient (including any salt or ester of the active ingredient),” § 355(j)(4)(D)(v) — to refer to the universe of all subsection (b) applications. Thus, the ambiguity of reference in the phrase reflects the possibility that it can refer to the latest subsection (b) application as well as all prior subsection (b) applications. Although our dissenting colleague asserts the language has a plain meaning, he does not explain why we are incorrect in perceiving this ambiguity.5
III.
Since we have discovered nothing in the legislative history that could be thought to embody the express intent that the language does not yield, we pass on to the second step of Chevron and ask whether the government’s construction falls within the bounds of reasonableness. We think it does not — at least its primary argument presented before us does not. The “reasonableness” of an agency’s construction depends on the construction’s “fit” with the statutory language as well as its conformity to statutory purposes. We cannot agree with the government’s unconvincing attempts to employ the “including” clause to cover all possible permutations of active ingredient. This case differs from instances where an “including” clause is designed to merely illustrate a few examples of the general category, see, e.g., Puerto Rico Maritime Shipping Auth. v. ICC, 645 F.2d 1102, 1112 n. 26 (D.C.Cir.1981). It is simply not plausible to read “including any salt or ester” as merely illustrative, to mean including any form that eventually produces the same active moiety. In other words, we reject the government’s interpretation as linguistically infeasible even though logically possible.
Once we reject the agency’s interpretation of the statute as unreasonable it does not follow that appellant’s competing construction must be adopted. Even if we thought appellant’s interpretation were reasonable we could not accept it if we perceived still other possible reasonable constructions. It is, after all, for the agency to make the choice between such alternatives. But in this case we do not believe Abbott’s suggested interpretation is reasonable.
Abbott argues that even if divalproex sodium is a salt of valproic acid,6 the stat*989ute precludes the FDA from treating the two chemicals as the same active ingredient because of the sequence of applications. To be sure, Abbott’s interpretation, at first reading, is the more obvious linguistic construction of the statute — that active ingredient refers in this case to dival-proex sodium. The parenthetical phrase (“including any ester or salt of the active ingredient”), according to Abbott, refers only to the original drug, the active ingredient “of which has been approved.”
Abbott’s interpretation then, unlike the FDA’s, is possible linguistically but fails to serve any conceivable statutory purpose. It would mean that if an original drug application has an active ingredient in the form of a salt, a drug company cannot obtain extended protection by merely filing a new application for a drug with an active ingredient in its non-salt form, but if it does the reverse, as in this case, it can. That construction appears to be farfetched because it is not consistent with any legislative goal: Abbott can advance no hypothetical reason why Congress (or indeed any of the interest groups) would have wanted the degree of protection a drug received to turn on this variable sequence.7 Abbott’s reading promotes neither the interests of the research-oriented pharmaceutical industry nor the generic drug industry in a rational way, producing instead a windfall depending on an accident of chemical nomenclature. We have not been offered any scientific, technical, economic or other explanation why Congress would intend the grant of a ten year market exclusivity to depend on the temporal sequence in which subsection (b) applications were approved and the counsel could not suggest any when pressed on this point. It appears, then, entirely serendipitous that in this case Abbott sought an application for the salt of valproic acid after first applying for valproic acid. Thus, under Abbott’s interpretation, if a pharmaceutical manufacturer developed two usable drugs, one a salt of the other, he could gain extra protection by applying for approval of the acid first, followed by the salt, but not under the reverse sequence.
If Congress truly intended the sequential distinction that Abbott urges, it would have drafted (it would have had to draft) quite different language in a different structure such as
If an application submitted under subsection (b) of this section is for a drug which contains an active ingredient that is also contained in a previously approved drug, then.... For purposes of this section an active ingredient in the previously approved drug includes an ester or salt of an active ingredient in the previously approved drug, but in the event that an ester or salt of the original active ingredient is filed as an active ingredient of the subsequent NDA, there is no disqualification.
That Abbott’s interpretation passes first reading suggests to us that Congress did not focus on the reverse sequence of application; not that it intended to address a sequential distinction. Ambiguity of language typically appears when the draftsman does not have in mind the manner in which the norm set forth in the statute will apply to a particular situation. We think that occurred here; Congress did not directly address the “precise question at issue,” Chevron, 467 U.S. at 843, 104 S.Ct. at 2782, and therefore the FDA (not the judiciary) is entitled to place its reasonable construction on the ambiguous statute.
* ‡ * * * *
Accordingly, we remand this case to the district court with instructions, in turn, to remand to the agency. We hold only that *990the statute is ambiguous; the reliance the government places on the including clause is an unreasonable construction and that Abbott’s interpretation of the statute is also unreasonable. Beyond that, we may not proceed since we have no authority to place a construction on the statute that the agency has not offered.
. If an application (other than an abbreviated new drug application) submitted under subsection (b) of this section for a drug, no active ingredient (including any ester or salt of the active ingredient) of which has been approved in any other application under subsection (b) of this section, was approved during the period beginning January 1, 1982, and ending on September 24, 1984, the Secretary may not make the approval of an application submitted under this subsection which refers to the drug for which the subsection (b) application was submitted effective before the expiration of ten years from the date of the approval of the application under subsection (b) of this section.
(v) If an application (or supplement to an application) submitted under subsection (b) of this section for a drug, which includes an active ingredient (including any ester or salt of the active ingredient) that has been approved in another application under subsection (b) of this section, was approved during the period beginning January 1, 1982, and ending on September 24, 1984, the Secretary may not make the approval of an application submitted under this subsection which refers to the drug for which the subsection (b) application was submitted or which refers to a change approved in a supplement to the subsection (b) application effective before the expiration of two years from September 24, 1984.
21 U.S.C. § 355(j)(4)(D)(i) and (v) (emphases added).
Both (D)(i) and (D)(v) apply to the transition period under the statute, from January 1, 1982 to September 24, 1984, during which the ten year and two year exclusivity periods apply. Sections (D)(ii), (iii), and (iv) control the applicable exclusivity periods after September 24, 1984. Because the instant case arose during the transition period, only (D)(i) and (D)(v) are at issue here. We note that the same language is used for determining the applicable periods of exclusivity for drugs introduced after September 24, 1984 — the only difference in the statutory scheme is the length of the exclusivity periods.
. The FDA based its decision about the chemical relationship between Depakote and Depakene in part on an affidavit submitted by Dr. Robert H. Wood, of which Abbott was not given notice.
. Of course the district judge’s construction of the statute is not entitled to deference as is the agency’s — if the agency's interpretation meets the Chevron standard.
. Subsection (b) applications are new drug applications for original drugs, while subsection (j) applications are generic drug applications which face reduced barriers to FDA approval. The length of market exclusivity simply determines when a subsection (j) application is allowed, which is determined by reference by the new drug application for original drugs, i.e., subsection (b) applications.
. We note that the Act includes other ambiguities hinging on unclear references. See Eli Lilly & Co. v. Medtronic, Inc., _ U.S. _, 110 S.Ct. 2683, 110 L.Ed.2d 605 (1990) (structurally unclear whether "a Federal law which regulates the manufacture, use, or sale of drugs" applies to individual provisions regulating drugs or to the entirety of any Act which has parts regulating drugs). In that case the Court had to pick its way between a more natural reading that yielded an implausible meaning, and an interpretation that gave the unclear phrase "a meaning it simply will not bear.” Id. 110 S.Ct. at 2687.
.The FDA found that the divalproex sodium is a salt of valproic acid but Abbott continues to deny it. This is partly a substantive chemical question and partly a procedural issue — -whether the agency improperly relied on an affidavit *989not brought to the attention of Abbott. Since we are remanding in any event to give the agency an opportunity to parse the language of the statute again, the agency might choose to adopt a procedure that would obviate any need for us to consider the affidavit issue if it remains relevant to the agency’s decision.
. Abbott’s brief did refer to a district judge's speculation that deriving ester or salt from an acid requires a greater effort than the reverse process, see Glaxo Operations UK Ltd. v. Quigg, 706 F.Supp. 1224, 1229-30 n. 12 (E.D.Va.1989), aff’d, 894 F.2d 392 (Fed.Cir.1990), but the district court offered no support for that notion and at oral argument Abbott’s counsel, when asked, did not put forth that hypothesis.