Counterfactuals

Lewis's account of a subjunctive conditional is that "if p were to hold, q would hold" is true if and only if

1. p holds in no worlds; or
2. some world in which both p and q hold is closer than every world in which both p and not-q hold.

But, plausibly:

3. No natural concept has a disjunctive account.
4. The concept of a subjunctive conditional is natural.

Therefore Lewis's account of a subjunctive conditional is not right. (Fortunately, there is a non-disjunctive modification of Lewis's account that has additional merits—see this paper and a reference therein.)

Is this a good argument against Lewis? I am not sure. If no disjunctive account of a natural concept is good, then no account of a natural concept in terms of the material conditional is right. Do we want to say that?

And, after all, the concept of a disjunction seems itself quite natural to us (certainly, much more natural than the connective nor, and probably at least a little more natural than nand). However, this answer may equivocate on "natural". Subjunctive conditionals, perhaps, are natural not in the sense that they are natural ingredients of our language, but in the sense that, unlike disjunctions, they are appropriate ingredients to have inside laws of nature.

[Edited]