The only way anyone gets the pretty flat graphs is by measuring in an anechoic space or taking a tiny slice in time before room reflections come in,
Measurements of the Mackies were taken in our anechoic chamber.
neither of which are represenative of what you will actually hear.
Actually the people that use these for mixing have them at the end of their mixing board aimed right at them (typically), so the on axis response is very representative of what they will hear.
Nothing is neutral or flat in real world practice. I just don't think the phenomenon exists, and I wonder if it is even desirable.
I have measured many speakers that are very accurate. We won't even release a speaker that doesn't fall within a +/-2db range and some of my work easy falls within a +/-1db range.
Is accuracy desirable? Certainly.
If flatness correlated 1:1 with sonic satisfaction then everyone would be installing wire floors and 3 foot fiberglass wedges in a barn. But what IS flat?
A smooth frequency response is only part of the picture. I have heard speakers with a flat (or fairly flat) frequency response fall short in other areas. A speaker might be good in other areas but have a roller coaster frequency response and be unusable or un-listenable.
Still, a smooth frequency response is a vital display of accuracy.
The Mackies were just outside of +/-3db with some rather rough peaks and dips that were a little too close to one another.
but I wouldn't doubt for a second that if I did the same test in my own room it would be the rocky mountains. Bare drywall will do that to a fella.
Are you really listening in a room with bare drywall?
I'd rather we drop the on axis FR analysis and use an average of 0, 15º, 30º, 45º & 60º (weighted if you prefer) it would be much more correlated with what we hear. Making a speaker flat on axis doesn't mean all that much if we listen off axis.
Actually the more valuable information would be to see on axis, 10, 20, 30, 40, and maybe even 50 degrees off axis. In room response measurements in an average listening room are helpful as well.
Another big one is vertical off axis responses. You can have good horizontal measurements but have a huge dip in the vertical off axis and wind up with a dipped response in the room response measurement.
If they sound good, who cares how they measure?
The recording engineers that are using them for mixing. That's who.
They may want one instrument to have the same output level as another that is in a different frequency range. They balance it to their ears using a speaker with a rough response, but then when it is played back on a more accurate speaker then they realize that one instrument is over twice as loud as the other was.
I use to have a local recording studio bring over their recently mastered CD's to listen to them in out listening room. Then they'd go back and remix, and come back again...