it's still scripted, pre-layed out, no games give you that sort of limitation during game play, why would you test under those conditions?
Well think about it this way. AnandTech, FiringSquad, Beyond3D, Hexus, HotHardware, Tech Report, etc have all used so-called 'canned benchmarks' for nearly a decade. And when has [H]ardOCP's conclusions ever disagreed with them? Sure the numbers come out a little different here and there, but the conclusions are always the same.
The term 'canned benchmark' vs 'real world benchmark' is essentially just marketing bullshit perpetrated by the [H] to make it sound like the [H] is somehow special. When in the end it all doesn't matter, the conclusions are the same. If you don't believe me, then go read the last 6 or 7 years worth of video card reviews and see for yourself.