So I realized that I was overlooking a thing in evaluating the math on the -5/+10 feats: I was overlooking the opportunity cost. Which is to say: If you didn't get that feat, what would you get instead? So the comparison isn't between using a -5/+10 feat and not using it. It should be, at least fairly often, better to use it than not to use it.
I think the right comparison is probably against taking +2 to your attack stat. So it should be compared against +1/+1, probably. Because the feat comes at the cost of an ASI. I suppose that becomes less true if/when you get the stat to 20, but if your stat isn't 20 yet, the -5/+10 feat is coming at the expense of increasing that stat.
I still think the feats are pretty good, but it occurred to me that all the modeling I saw was assuming that the alternative was just "don't have that feat", rather than "have something else instead".
I think the right comparison is probably against taking +2 to your attack stat. So it should be compared against +1/+1, probably. Because the feat comes at the cost of an ASI. I suppose that becomes less true if/when you get the stat to 20, but if your stat isn't 20 yet, the -5/+10 feat is coming at the expense of increasing that stat.
I still think the feats are pretty good, but it occurred to me that all the modeling I saw was assuming that the alternative was just "don't have that feat", rather than "have something else instead".