There are two ways to look at accuracy: accuracy of the dyno itself, and the effects of introducing other variables outside the measurement system that can affect what the car is actually doing. As long as the dyno is properly used and calibrated, then all dynos are accurate for the conditions. The pitfall is trying to compare results across different platforms.
......
To the extent possible, you want to run multiple tests within a consistent shop environment, including similar ambient temperature and pressure. The more rigorous the setup, the more exacting the dyno calibration (on those load-sensitive dynos relying on brakes and absorbers), and the more controlled the overall shop environment, then the more repeatable the results, and the more suitable the facility is for measuring the effect of small incremental changes. Still, once you get down in the 5 percent range on the average chassis dyno in the average facility, it is difficult to evaluate whether the change is real or due to environmental transients.
Read more:
http://www.hotrod.com/techfaq/hrdp_0...#ixzz2GxRqCIRS