You still haven't shown that the increased risk of danger outweighs the aggregate time savings. It's a big consideration when looking at things like, the value of screening for unlikely outcomes.
3.5 minutes/trip saved * 2 trips/day * 200 working days per year (probably more) = 1400 minutes saved per year per person.
working-age population of US (2012) = 243 million.
Employment-to-population_ratio = 67%, so 162 million.
Let's say that half drive, or 80 million.
So, 80m * 1400 minutes = 112 billion minutes a year saved by speeding or 1.9 billion hours. At a wage of $10 per hour, that's $19 billion, which is a bit ahead of the $15b loss of value per your number, but I'm sure our confidence intervals overlap.
112 billion minutes = 213000 person-years of being stuck inside a car saved by speeding. With a US average lifespan of 79, that's 2700 total additional lifespans spent in a car by losing those 3.5 minutes.