Curious statistics,or perhaps I misunderstand, but how do you get accident frequencies v alcohol consumption from dead drivers.
Alcohol consumption would be ascertained by taking a blood sample post mortem, counting the bodies would be simple.
I don't remember the details of the statistics as it was around 20 years ago, but I remember the graph shape and the conclusions drawn from it. Consider it anecdotal evidence if you will, but I think the concept is sound and should be understood by most.
I suspect they had a good idea of what the frequency of fatal accidents was for sober people and that was used as a baseline. They could then look at the frequency of the accidents for dead drivers with non-zero alcohol levels, and compare that with what would be expected. The graph came from that.
I guess the bottom line is that at the legal limit as it stands you are measurably impaired and you should not be on the road. A one time thief should not be a police officer ever, a one time child molester should not be a schoolteacher ever, and a one time drunk driver should not be an airline pilot ever, imo of course.