"The point is that the guy's BAC was certainly above 0.15"
That is nothing more than an outright guess for you also do not know what he drank or when.
"Accident rates as a function of BAC are pretty flat below about 0.12, when they start going up very sharply."
"An April 1999 study of all fifty states conducted by the National Highway Traffic Safety Administration (NHTSA) confirmed MADDs assertions. It compared states with .08 BAC laws to states with .10 BAC laws before and after the laws were passed. The study found that states that passed .08 BAC laws reduced the involvement of drunk drivers in fatalities by 8 percent."
http://www.enotes.com/drunk-driving-article/
I think matters. Proof being in the pudding and all.
The linked page did not link to the study, so it's unclear what methodology was used to conduct the study. Without knowing that, it's impossible to offer any meaningful interpretation of the numbers. Certainly there are many ways such "studies" can be constructed to yield "results" far different from what the actual data show.