I can only speak for my dyno as far as accuracy at low rpm. It can measure accurately but if the engine has a " big cam" the engine might not run well at low rpm and if the the engine doesn't run well it might not repeat well. Not having repeatability makes it difficult to evaluate changes.
Racers have been using dynos to determine safe starting point tunes for race cars for probably as long as there have been dynos.
If you take two identical motors unfamiliar to the tuner and race them side by side, I think the motor with a well sorted tune off the dyno has a better chance for success than a motor that does not. Not to say that eventually the other motor couldn't be tuned to run as fast.
