Mileage testing, as done by the magazines, is inherently imprecise. First they use different drivers, and even the same drivers will vary from day to day. Secondly the road conditions are different, as is the wind and weather. Are they all using the same sort of fuel monitoring systems or are they just depending on filling up the tank. That will vary depending on the fuel temperature. Do they use the same tires? What about tire inflation pressure? Lots of variables.
If you had a robot driver that could drive each truck exactly the same way over the same road and with the same weather and other conditions, then you might see somewhat reproducible numbers. You'd have to repeat the test with the same truck multiple times in order to get an average number. In the lab under highly reproducible conditions the certifying agencies require that a test be run 40 times in order to generate usable statistics.
