Wait, what? The analogy works exactly; you're just assuming a priori that the bit you think doesn't fit actually doesn't fit. The analogy logically goes that if it's wrong to point a gun at someone regardless of whether you think it's loaded because it might be anyway and that would be Very Bad, it's also wrong to proposition women in elevators regardless of whether you think they'll accept because the situation where they don't would be Very Bad.

No, the position of skepchic et al is (as best I can tell) that it would be no big deal if the propositioner were hot, at least to the extent that e.g. she wouldn't be making a blog post about, "hey, this guy I was really into asked me to his room [different terminology because she likes him] when we were on the elevator, and we had a great time, BUT YOU SHOULD NEVER DO THAT and don't take my acceptance on this occasion as an indication that it's okay, and I made this clear to him and informed his friends that he did something obviously very dangerous and which they should not repeat."

Now, you may have a point that there is a similarity between the two ("danger justifies erring on safe side as a rule"). However, there is a more important difference between how they're handled: specifically, that correct guessers on propositioning in elevators are rewarded, while correct guessers with gun handling are still punished (even if it's just a verbal rebuke). Yes, I suppose you "should" err on the safe side in both case, but as a practical matter no one is anywhere near giving a damn on correct guesses in one case, while they are very concerned in the other.

And this has fundamentally screwy incentives effects.