How a dyno is operated can make quite a difference in HP readings.
From my life-long best friend, who sometimes operates the (car engine) dyno at a race engine shop in Tiffin, Ohio, and from my own experience at a circle track engine shop in Phoenix, AZ (where they have a very nice dyno room), I've learned that the way the operator loads the engine during a "pull" can very the results.
Generally, a dyno operator usually ends up overloading the engine during a dyno run. That is, he sets the water brake to hold the engine at the various rpms a little too long. This does several things, including showing more HP on the print-out, and causing the engine to seem a little lean. Circle track guys are notorius for doing this. Drag guys aren't much better, either.
A faster accerating rate should be used, as it approximates the real world more than a heavily loaded, slow accerating rate does. Some engines rev up faster than others, for a variety of reasons, including - lighter crankshaft, flywheel and other moving components - or more spark timing advance (more is almost always better!). This faster revving rate, "a good thing", would not be obvious with an overloaded run.
Another problem with dyno runs can be the operator's set up in the computer that records the various output of the sensors. That computer can be programed to fudge the numbers by whatever amount the programer wants. Ten percent is not uncommon to add to the stock readings.
NOTE: I'm not talking of adjusting the computer results to an STP day here. (STP means Standard Temperature Pressure.) Doing that type of adjusting gives a baseline so that guys can compare their results meaningfully with others.
All of what I'm talking about here is for crankshaft HP testing. As DanB notes above, a chassis dyno will show lower readings because of driveline losses between the crank to the wheel.
