sorry for jacking the ched, but here's what i found on inertial and steady state dynos:
Typically a steady state dyno will give results of up to about 20% less than inertia types. Here is an example of how different results occur. You have a bike that shows a maximum of 100HP on an inertia dyno and xxHP on a steady state dyno. Now we lighten the crankshaft and flywheel, fit a lighter rear wheel, fit a lightweight chain and some alloy sprockets. We run the bike on the steady state dyno and it still shows a maximum of xxHP. We run the bike on the inertia dyno and find it is now producing 105HP. These modifications didn't actually make the engine produce more horsepower just as the brake dyno shows. So why does the inertia dyno now say it is producing more horsepower? This is because the inertia dyno gives a true representation of what the "road" sees. Of course the engine isn't producing more horsepower, but there is more horsepower available to accelerate the bike because less power is needed to accelerate the crankshaft, chain and sprockets and finally the wheel. Because less power is need to accelerate these things more is available to accelerate the bike, and it will accelerate faster on the road. The inertia dynamometer calculates horsepower from how fast its drum is accelerated, therefore in this example the bike engine was able to accelerate the drum more quickly after the modifications so more horse power is available at the rear wheel to accelerate the bike on the road. In my opinion steady state dynos are good for tuning tractors that are going to run at 1500-rpm day in day out. To tune an engine for the type of riding I do, or bush riding or racetrack riding I think an inertial dyno is more than sufficient. How often do you care how much power you're making at a steady throttle? When I'm out having fun, the engine is rarely at a constant speed, it is either constantly accelerating or decelerating.
taken from
http://wotid.com/content/view/17/36/
according to this, Audi's dyno should be better for simulating real road conditions, since it should give results that 'the road will see'