As a former dyno design engineer, you are totally correct. This approach is especially important for chassis dynos, where there are tons of influences on results (try letting 2psi out of your tires between runs!) and wacky estimates for crank power and torque, loosely based on numbers generated at the tires.