Unfortunately due to all the interactions, as you say a dyno operator/tuner can get just about any number they want from a car. Some do it with full knowledge and fore thought in order to sell a specific product or service. Others not so well educated on how all these inputs effect the output, do it out of simple ignorance. They don't even have a clue they are cooking the numbers. Last are the ones that due to pressures from miss-educated consumers bow to the pressure of the market place, and tweak the setup to give numbers comparable to what the consumer expects. Not because he wants to, but because he must in order to stay in business. Especially in the case of a tuner who's work is frequently judged on peak power not percentage of change in peak power, its a question of survival to produce "commonly expected" results on the dyno.
It would be interesting for someone to take a car and publish a comparison showing the wide range of numbers that they can get from an unchanged car with simple and not so obvious changes in the dyno setup. Ie tie down tensions, tire pressures, tire types, oil temps, gear box temps, changes in lube oil and gear lube viscosity, over cooling the intercooler, under cooling the intercooler, pressurizing the dyno cell with ventilation fans, etc. etc.
The real problem here is that the buying public is for the most part totally ignorant about dyno variablity and actually believe a dyno printout is a hard and fast "proof" of a specific power level. Until the average consumer quits looking at a dyno output as a hard and fast number, and starts seeing it as a relative number I suspect the problem will not go away.