Theres a lot more to dyno testing a car than just trying different gears, theres also ramp rates ( the rate of acceleration that the dyno allows) and the amount of load applied.
If you do get a different power reading in 3rd gear to 4th gear then theres something wrong with the way the dyno run is conducted, maybe the car didnt hit full load in one gear but did in the other, maybe a wheel bearing is worn or a brake is binding, this would give a different (lower) reading in a higher gear as the rotational speed of the drivetrain is higher and the restriction would give greater power losses for the given RPM.
There are far too many variables when you try and compare one dyno to another, this is always an issue when customers rely on the flywheel power figure, the big difference that I see isnt in the true power that a chassis dyno (drive on dyno) measures i.e power @ the wheels, its when you compare dynos based on the flywheel power, which after all is nothing more than a 'good guess' that the dyno works out. This isnt the case in the USA where power @wheels is more commonly used.
We have tested a car recently that made 387bhp at flywheel on an old coastdown type dyno (these measure transmission power losses by coasting down from peak rpm to 0rpm at the end of the power run and measuring the resistance during this coastdown, working this out as a percentage of power @wheels and adding it back on to give a flywheel figure) this car made just 302bhp at the wheels, on our dyno it made 349bhp at the flywheel but strangely 298@ the wheels. So a difference of just 4hp at the wheels but a massive 38hp at the flywheel!
So which dyno was right? Well both gave wheel power figures within less than 1.5% of eachother, but one dyno had massively compensated for the flywheel figure, maybe ours had undercompensated? Who knows?
My advice is to always ask the dyno operator for the bhp at both flywheel and wheels and use a modern dyno that doesnt use coastdown to generate its power figueres, then you cant go far wrong. If your result has been 'fudged' then you can tell by comparing the power@wheels to the power@flywheel.
Theres a lot more to it than that but ill leave it there
