dynos read wheel horsepower, auto manufacturers rate thier vehicles at crank/engine horsepower. by the time you factor in parasitic loss through the drivetrain, the actual horsepower that is hitting the street is much lower than what a manufacturer prints. for example, a protege from the factory was rated at 130hp but most dynos put the car at barely 100 to wheels. this is where you will see abbreviations such as "whp" - wheel horsepower - and "rwhp" - rear wheel horsepower- etc etc. this dyno was also done on a mazdaspeed6, where the all wheel drive system will show more parasitic loss than a two wheel drive setup. you also have to take into account the type of dyno being used - a dynodynamics system typically rates lower than a dynojet or a dynapack style system, so on and so forth. these variations along with personal beef will often lead people to cry foul or cry BS when a dyno is posted, regardless of how accurate it may be as there are many factors which can affect a dyno result. the dyno's correction factor, the RPM at which sampling started, the climate - all of these and more are factors which can lead to dyno variation between two cars and even dynos of the same type and brand.
whats important when modding is not the number itself, but the consistency of difference between the baseline and the modded result. so if one car dynos stock at 200 and another 220, with the particular mod providing gains of 220 and 240 respectively, give or take a few, this shows that the mod provides consistent gains vehicle to vehicle.
for the record, on cobb's forums, the mazdaspeed3 dynoed stock at 215whp. i have had a canadian customer tell me his mazdaspeed3 dynoed in the same range on a different system.
there is also the case of factories overrating and underrating thier own vehicles from the factory. dodge claimed thier srt4 neons initially had 215 hp, yet when dynoed - even stock - the vehicles would show closer to 230 or 240 to the wheels! GM had the same 'issue' with saturn ion redlines/cobalt ss supercharged where they would dyno higher than rated. mazda unfortunately has a recent record of overrating thier vehicles. the rx8 was initially rated at 255 hp, yet independent testing showed that the number at the engine was around 238 hp. the same happened with the miata. it was worse on the rx8 automatics, where factory rated horsepower was adjusted to below the 200 mark. the severity of this error on mazda's behalf led to buyback offers and free maintenance compensation to thier customers.
sorry for the lecture, not even a fully comprehensive one at that, but it explains for those not aware why dyno numbers don't match the brochure - basically two different numbers at work.