In practice, less than we'd like. I would apply the same caveats you would looking at a speaker frequency response measured in someone else's living room. The acoustics of the tiny "room" your head/ears create in conjunction with a given headphone (to say nothing of the subjective qualities thereof) could be quite different. But measurements made on the same rig are useful for gauging the overall tonal differences between two cans... to a point - for example, to my ears my AKG K371s have some very prominent peaking near 6khz I don't see on any measurements made on the GRAS artificial heads.

I've yet to see compelling evidence that at normal listening volumes "dynamics" is something separate from frequency response and your usual illusory differences that have nothing to do with the actual sound.