A chassis dynamometer's first function is to measure the amount of mechanical energy available at the rear wheels. This is what a DynoJet does very well. It's probably a good bet that you can turn off all the simulation crap on a Mustang Dyno and get the same exact numbers.
My problem with what I've been reading is that it sounds like tuners are misusing the intent of the simulation. I agree with turbolx's idea of simulating load for the purpose of finding issues with pinging and such. What I disagree with is the idea of measuring horsepower during a simulation. The reason you're getting lower numbers is that you have two sources of mechanical power pushing against each other. The first is the REAL output of the motor. The second is the SIMULATED output of the dyno itself. It sounds like you are basically reporting the net horsepower between the two. That's a worthless number since you're mixing actual data with simulated data.
OR (maybe turbolx can answer this) the Mustang Dyno knows exactly how much force it is applying against the engine and can make adjustments to the drive wheel horsepower accordingly. That would be better, of course, but if the simulation isn't accurate and simulates too much resistance, you're potentially tuning out power that you would have otherwise had.
If I would tune my car on a Mustang Dyno, I would at the very least expect that my final horsepower numbers come from the drive wheels and that the simulation stuff is turned off at the time.
As far as turbolx's comments about air/fuel ratios being different under certain loads I would agree and the computer does have tables based on varying loads at different RPMs. I hope, however, he wasn't implying that there is a way that he can tune that without burning a custom chip or using a completely programmable ECM.
------------------
351W 89 Mustang GT Convertible