I just ran into an article that compares the 2 most common types of dynos. This isn't the whole article, it's pretty long, but if you want the whole thing I'll email it to you.

Dynamometer Results - Fact or Fiction?
This discussion revolves around chassis dynamometer's and is intended to be informative and thought provoking. There are two types of chassis dynamometers on the market, inertia and loading. An inertia dynamometer (such as DynoJet) does not measure torque, but measures acceleration. A loading dynamometer applies resistance that is measured (using some type of strain gauge.)
The most often heard discussion is that what factor can be applied to rear wheel horsepower to reflect crankshaft horsepower. This is where we need to understand how the rear wheel horsepower number was derived. Since the DynoJet seems to be widely used and numbers quoted are those from a DynoJet, we are going to use them as our inertia dynamometer example.
First it is important to have an understanding of how DynoJet gets their horsepower numbers. Power in mechanical terms is the ability to accomplish a specified amount of work in a given amount of time. By definition, one horsepower is equal to applying a 550 pound force through a distance of 1 foot in one second. In real terms, it would take 1 HP to raise a 550 pound weight up 1 foot in 1 second. So to measure horsepower, we need to know force (in pounds) and velocity (in feet per second). Dynojet's inertial dynamometer measures power according to the terms just described. It measures velocity directly by measuring the time it takes to rotate two heavy steel drums one turn. It measures force at the surface of the drum by indirectly measuring it's acceleration. Acceleration is simply the difference in velocity at the surface of the drums from one revolution to the next. The force applied to the drums is calculated from acceleration using Newton's 2nd law, Force = Mass * Acceleration. Since the mass of the drums is know and acceleration has been measured, Power (horsepower) can now be calculated. Torque is then calculated using the horsepower number: Torque = Horsepower * 5252 / RPM.
Once they have these numbers a series of correction factors are applied, some made public, some hidden as proprietary secrets. The public correction factor is the SAE correction factor. This formula assumes a mechanical efficiency of 85%. The formula used is: Where: CF= 1.18 * (29.22/Bdo) * ((Square Root(To+460)/537)) – 0.18. To = Intake air temperature in degrees F, Bdo = Dry ambient absolute barometric pressure. This correction factor is meant to predict output in varying atmospheric conditions and is a +/- 7%. The proprietary correction factor is supposed to reflect the loss of power from the crankshaft to the rear wheels.
A Loading Dynamometer applies resistance to the dyne's roller(s) , typically using either a water brake or a current eddy brake. In either case, the amount of force is measure using a strain gauge. The measured force is torque which is a real, indisputable measurement of the actual output at the wheel. Horsepower than can be calculated: Hp = Trq * 5252 / RPM.
A Dynamometer can only measure actual power at the output location. Actual power produced AND delivered by an engine will be highest if measured at the crankshaft, lower at the transmission output shaft and even lower, but more meaningful, still, at the rear wheels. The power that you use is the power at the rear wheels. Some Dynamometer companies add to measured rear wheel power readings a factor that is based on ESTIMATED rear wheel power losses (under what power conditions? 3.0 ltr.? 5.0 ltr.? Under coasting conditions? with a 185/70/15 radial tire? a 335/35/18 radial tire? New heavy radial tire vs. worn old, light, racing tire? Who knows?) In short, there is NO meaningful "average" tire to get a correct rear tire power transmission loss measurement for all cars - so obviously, unless they actually measure the power lost in the rear tires, under driven load conditions, NO dyno company should BE ADDING incorrect power figures into the measured power. It's simply wrong. The fact that they add varying amounts of power to the actual, "true" amount of power delivered and measured to the surface of the drive roller creates a situation that makes it an onerous task to compare power figures from different brands of dynamometer systems. On simple inertial dynamometers, some (most) companies use an average for the inertial mass value of the engine, transmission, driveshaft, axles and rear wheels. This is saying that a 4 cylinder, 2.0 ltr. Porsche 914 has the same rotating mass and same rear wheels as a 8 cylinder, 5.0 ltr. Porsche 928 S+4. This simply is not so and wrong.

 

TTT for those that didn't see it earlier.