Allow me to make a correction.
That category is 'killed in alcohol related accidents'.
What does that mean? Is there a difference?
If I am a sober cab driver that picks you up at a bar to take you home and I get hit by another sober driver, it gets labeled as 'alcohol related' because you (even though you weren't driving) had consumed some alcohol.
40,000 people a year die in car accidents and even taking the 15,000 as gospel, 25,000 get killed by sober STUPID drivers. Maybe cars and people are the problem, not alcohol. ;-)
Now, let's compare the miles driven by drunk drivers to the miles driven by sober drivers. And then ratio the fatalities. I imagine it would be an order of magnitude higher, at least, for the drunks.
And this gets back to my main point - we do have risk threshholds for taking sanction before harm happens. I agree the .08 is too low - it should be back up to .10, and cops should look for actual driving impairment instead of having checkpoints. However, it does not change the nature of the overall risk factor. And since pot has a much lower risk factor than alcohol or hard drugs, it's inane for the feds to criminalize it.