I tried to find corroborating sources and couldn't, so take this for what it's worth, but Claude seems to think the cutoffs are based on epedmiological evidence for health impacts at various BMIs, which, even if that _wasn't_ how it was arrived at, seems like a better option.