Just because of the ease of computation and discussion.
Because it's easier to say "if you slow down X MPH, you will increase your MPG by Y%" than to rattle off "if you slow down by 1.3X ^ 1.08 MPH, your MPG will increase 0.89Y^1.17 times 0.162 MPG".
You are absolutely correct that the MPG difference between 50 and 65 is much smaller than the MPG difference between 65 and 80 MPH, but I'd prefer to not carry a calculator to figure out what's gonna happen when I want to speed up or slow down 5 MPH.
Besides, I enjoy doing linear regression. I'm pretty much done doing anything more complicated...