Here's a little more information about how dyno's calculate horsepower. I wrote this for another site some time ago.
The basic reasons for the difference between a DynoJet and Mustang Dyno, is in part on how the power is calculated, and partially due to the amount of "load" the vehicle "sees" during the pull.
DynoJets measure horsepower and derive torque from that. Loaded Dyno's have a load sensor, and actually measure the torque being applied to the drums (it's a little 5volt reference load sensor). Horsepower is then derived from that information. Torque is an actual force (like gravity), where as Horsepower by definition is a derivited of torque (work over time).
The load sensor data is then sent to the computer where it takes two constants, vehicle weight and Hp@50mph (aerodynamic coefficient used by the EPA) where the computer then calculates the amount of resistance to be applied to the vehicle (PAU force).
For an example of why there's a difference due to load, imagine this. It takes less power (however you measure it) to accelerate the mass of a 2500lb roller from 300rpm to 400rpm in five seconds than accelerating the same roller (2500lbs) with 550lbs of resistance via load cell from 300rpm to 400rpm in the same amount of time (5 seconds). That's why if there's a car that made 425rwhp on our Mustang MD-1100SE dyno, it will be faster than a car that made 425rwhp on a DynoJet.
As far as the timing example that somebody brought up, I've found that to be true as well. Sometimes the cars are faster with 1 to 2 degrees less timing than they see on the Dyno. Best was without being at the track is to check the plugs. That's why, no matter how good the dyno is, the opperator, or the tuner is, it will really only get you about 95% there. The last 5% is going to be needed to be done at the track.
Because of these reasons, it's like trying to compare apples to oranges. As far as the "10%-15%" difference, we've found that to be untrue. In our tests it doesn't seem to be a constant at all. At 390rwhp on our dyno, the same car makes about 405rwhp on a local DynoJet. At 422rwhp on our dyno, the same car on the same DynoJet made 458rwhp. I've witnessed cars seeing close to a 80rwhp difference between a Mustang Dyno and a DynoJet at 600rwhp (Mustang MD-1750SE), although I cannot verifiy that with our dyno. More recently, I've found that a 2002 Vortec S/C Z06 that made 550rwhp on our Mustang Dyno, made 645rwhp on a DynoJet over the summer.
We're fairly conservative, and hence tell our customers that the difference is closer to 6-7%, but as you make more power, the difference increases as well. You must remember, Dyno's regardless of the type are tuning tools, and are in no means ment to tell people how fast their car is. Now which one is more "real world" is a totally different question. I like to explain it like this..... If you drive your car in a situation in which you have no mass and you're in a vacuum, so basically if you do intergalatic racing in space, use a DynoJet. If your car sees gravity, and has an aerodynamic coeffecient, and you race on a planet called Earth, then use a Mustang Dyno.