Ted raises a very interesting issue about logic.  I don't recall any discussion of it in print, though no doubt people besides me have said things that have implications for it.
     Ted is right that my (tacit) policy for individuating arguments is somewhere in between the purely grammatical policy and the purely logical(-form) policy.  And he is right to ask why that should be, when each of the pure policies is straightforward and impurity seems ad hoc.  I will defend my impure policy, though.  (Let me begin by confirming that I do think the individuation of arguments and instances is a conventional matter.  The only considerations that bear on it are pragmatic ones.)
     As Ted agrees, the purely grammatical policy is a non-starter.  No one would ever think the boat-human argument was a proper instance of &-Introduction.  The reason is, that semantic-pragmatics (incorporating minimal charity) rules that "She" is equivocal as between the two premises.  We know that just by looking.  (Of course, there could be a context in which we would not read "she" as changing its reference, but it would have to be a very special one.)
     I have two objections to the purely logical-form policy, both rooted in the fact that ordinary philosophers do not know the precise logical forms of many sentences--not of any sentences, if any version of Montague Grammar is correct.  Linguists and linguistic semanticists don't know many either; there are precious few (if any) established results in semantics.
 First, take Ted's own example:

        (P1)I am carrying an umbrella
        (P2) If I am carrying an umbrella, it is raining.
        (C) Therefore, it is raining.

        (P1) An umbrella is being carried by me.
        (P2) If an umbrella is being carried by me, it is raining.
        (C) Therefore, it is raining.

As a matter of fact, it's quite controversial whether actives and passives have the same logical form.  Across the board, they certainly do not (our own Paul Ziff proved that back in the 1960s).  In Montague Grammar, even simple active-passive pairs like Ted's differ in logical form.  So, if we individuate by logical form, (at best) we don't know whether Ted's two instances are instances of the same argument.
     Second, and more to the point of the present issue, there are lots of context-sensitive locutions about which it's controversial whether the context-sensitivity is explained by shift of a hidden parameter.  As we saw, in my semantics for conditionals, thir context-sensitivity is so explained, but in Stalnaker-Lewis semantics it isn't.  The dispute between the two theories will not be settled any time soon.  Until it is, if we hew to the line of logical-form individuation, we won't know whether Dretske's argument and the standard counterexamples to Antecedent-Strengthening equivocate or not.  (And there are scads of other kinds of expression that provoke similar debate.)

     Ted writes concessively:

In other words, when we individuate by grammatical form, we have the benefit of easily and uncontroversially telling when an argument is an example of modus pollens, modus tollens, etc. Of course this is true only if we are conveniently ignoring arguments that use indexicals, homonyms, proper names, etc. Considering those cases, then, the idea might be that individuation by grammatical form (plus the exception clause) is better because (a) in cases without indexicals and the like, individuation by grammatical form is easier than individuation by logical form; (b) in cases with indexicals and the like, your view individuates the arguments in the same way as the view that individuates all arguments by logical form. So since your view is at least as good in some cases, and better in the other cases, it is better on the whole.

        (A) Individuate all arguments by logical form.
        (B) In cases without indexicals and the like, individuate arguments by logical form via grammatical form.  [I.e. when there's no indexicals et al., use the following rule of thumb: whenever you have the same grammatical form, you have the same logical form.]  In all other cases, individuate arguments by considering their logical form "directly" (whatever that may mean).

     Ted now justly asks why I or anyone should treat (L)-(W) as non-equivocal and a genuine instance of  Antecedent-Strengthening while still making an exception of the boat-human argument and dismissing it as a non-instance of &-Introduction.  Isn't my policy still ad hoc?  In fact, is it even really a policy, given that I've just picked and chosen my examples the way I felt like, rather than by appeal to any principle?  And as Ted says, any mere stipulating I do in this matter should accord with our actual practices in logic.
     Right; here's my principle:  Individuate by grammatical form, except when normal semantic-pragmatics rules that a term occurring (overtly) in the argument changes its meaning or referent between premises and conclusion.  That exception clause takes care of the boat-human argument, all equivocal use of pronouns, shifts of word meaning, etc.
     (Of course, normal semantic-pragmatics may not rule decisively.  It may only raise the question of equivocation on an occurring term and/or make equivocation likely to some degree.  In that case, we suspend judgment until we have disambiguated.)
     There will still be problem cases.  For example, my appeal to occurrent terms may fail to capture purely syntactic ambiguities:  "Visiting relatives can be boring / \ To eliminate the possibility of boredom, you must not visit any relatives."  I'll pass that by for now.
     What about Ted's concluding example?:

        (P1') Everyone made it to the party.
        (C') Therefore, every person from every nation made it to the party.

I agree it is and should be counted as invalid, and (as a matter of fact) I agree it fails by parameter shift, though not every semanticist would grant that.  But I think normal semantic-pragmatics takes care of it.  Any normal speaker/hearer knows that "everyone," especially when operating on a predicate indexed to something local such as "made it to the party," does not mean every person from every nation.  (And for that matter, the argument is flatly invalidated just by the fact that the conclusion entails the existence of nations, while the premise does not.)

     I don't claim to have settled this issue.  I think it remains open and interesting.  Thanks to Ted for taking the trouble to raise it.