"the reduction in fatal crashes by those who habitually drive severely intoxicated could easily account for the entire observed phenomenon."
I would offer that the study showed that reduction applied to both and it took a harsher revokation/suspension penalty to be used in combination.
I think it shows that both used in comination with each other reduced the deaths. It appears that you aim to say it was punishment alone that accomidated the reduction and that would be as untrue as to say that reducing from .10 to .08 casued the reduction alone.
Major problems require multiple pronged approaches to effectively deal with them.
I think the study(ies) show quite conclusive backing that this was effective. I would offer that further reduction to a .06 or lower would not have the same effect as the drop toi .08 did. But that is just imho.
Without knowing what specific measures were implemented when, and in what states, it's impossible to assign any meaningful cause-and-effect.
BTW, I would further suggest that if the state were to pass and enforce a law against driving with fuzzy dice hanging from the rear-view mirror, that would probably result in a reduction in the number of fatal crashes by drivers with such fuzzy dice. Such reduction, however, would not mean that such a law saved even one life.