The results from using an average of several measurements across a "window" has proven to me the worth of it. These results have been better sounding systems than for any other method I have used. I believe that the theory behind this technique is sound (excuse the pun).
This theory is that the power response (the total power output from the speaker into the room), is of high importance. The on-axis response is the most important, in my opinion, followed by the power response, as this interacts with the room and has a strong influence.
Speakers are most often positioned pointing either straight ahead or, more often, toed in slightly. Because of this, I believe that the off-axis response on the side closer to the listener, is possibly as important as the on-axis response, as this is the true first-arrival response.
Offset of the tweeter is also easily worthwhile. The resultant output is more linear in almost every way. At the same time, the average of the windowed measurements is much smoother. The sound is subjectively improved as well.