The BAC is a lot like the Air Force's cycle ergometry test of yesteryear. If the only thing that affected your heart rate was VO2-Max, it would be an accurate test. However, since your heart rate is affected by all manner of things, physical and psychological, it was a horrible test for accurately determining fitness.
If the goal is to reduce traffic fatalities, we need to talk about the actual ability to drive from a performance perspective. There are many factors which affect performance and, as such, performance-based testing would be the ideal. Since we don't have that, we whittle it down to a simple and arbitrary BAC number which does nothing for safety other than make people feel good.
The simple truth is alcohol affects people differently. Fatigue impacts/impairs drivers. Cold medicines, distractions, kids, you name it - all go to performance behind the wheel. There needs to be a performance standard, not an arbitrary number. Now that CO & WA have legalized pot, how do they measure impairment for driving? Is there a THC level above which you cannot drive? Or is it zero tolerance? If so, how is it determined? Is there an equally arbitrary baseline as there is for alcohol?
I've been told, but cannot find any reference to it, nor can I find anyone who will confirm it, that Canada used to have a drunk driving license back in the '50s-early '60s. Supposedly, you could take the test at up to .10 and if you passed, you got a note on your license saying you were legal to drive up to whatever you tested at. Sounds ludicrous today, doesn't it, but (if true) it speaks to the fact that alcohol affects people differently from a performance perspective.
Some industries whose employees regularly work with dangerous machinery are trying to find a performance-based test they can use to evaluate their personnel prior to beginning their work. The challenge with simple reaction-time testing or something similar is that people can 'sandbag' the test endangering themselves and others if they 'pass' while 'performance impaired'.
There's so much more to drinking and driving, but it's been boiled down to an arbitrary number that is far from uniform in its effect on any number of individuals. Truly, reaction time is a huge portion of safe driving and reaction time is nothing more than the ability to process information and respond to it. In the fighter-pilot jargon, it's the OODA loop. Folks with a fast OODA loop while driving are generally better drivers than those whose OODA loop is slower or less developed. If you think on this concept, you'll see why inexperienced drivers, as well as elderly drivers, are in more accidents - their OODA loops are either not fully developed, or are decaying. Folks who are just sick, even if not medicating, are not 'on top of their game' and therefore, unsafe behind the wheel.
There is the rationale that everyone is impaired after 1, 2, 4, 10 drinks. The real challenge is impaired down to what? If there was a performance standard, would it really matter how much you'd had to drink so long as you could still perform at the required level?
It's a rabbit hole, I admit, but it's far more palatable than some random and arbitrary BAC number.