"There are lies, damn lies, and statistics" (Mark Twain)
Substitute "loudspeaker measurements" for "statistics" and you understand my point of view.
After thirty years in audio I have learned to mistrust measurements of all kinds on audio equipment, at least as they correlate to listening quality. When I started out I ran a QC electronics laboratory (this was in the late sixties and early seventies) where one of my jobs was testing audio equipment, so I did a lot of it. And in those days the deficiencies of much gear were easy to measure. Most commercial receivers I checked for THD, IM and amplitude linearity had problems in all three areas, with bass rolloffs notable below 40Hz, often quite steep. This was also true of tape recorders (even the best open reels of the time) and of course, speakers. Speakers were typically attenuated at both extremes and often peaked in the "warmth" region--the "east coast" sound, or in the lower trebles--the "west coast" sound. Audiophiles merely decided between bright and lean, or harsh and peaky, then did their best to find gear that compensated for the problems.
When I went into speaker meanufacturing in 1977 I was not happy with what I measured and soon restricted mic time to transducer work, rather than finished systems. In transducer design it was essential to concentrate on diaphragm resonances and FR peaks/valleys which you could ameliorate by changing motor, suspension, and diaphragm. As materials improved so did drivers. THD was higher than in speakers than in amplifiers, particularly in the bass range, until I discovered push-pull bass, slot loading, and mass loading of passive radiators which could be tuned for lower distortion. By the mid 90's I was measuring THD below 2% on properly tuned speakers right down into the first octave.
With the advent of computer measurement programs and CAD even amateurs could design speakers that measured flat. Among hobbyists it became the fashion to see whose speakers could measure flattest. This led to a "linearity derby" with a narrow focus on flat on-axis amplitude response achieved by any means possible, including high order filters (whose ringing was not apparent in sine sweeps), out of phase driver wiring (which smoothed FR in the crossover region while introducing a completely amusical polarity reversal in the passband), steadily increasing Xmax (not a bad thing in itself except for increasing manufacturing costs), and finding a measurement "sweet spot", usually at 1m on axis that completely overlooked the fact that no user ever listened to speakers there or ever got the benefits of the 1" vertical and horizontal window yielding these fine results. Frequently, performance away from the "sweet spot" was poor, leaving audiophiles to wonder why "flat" speakers sounded so unsatisfying.
When we were building the Spica TC60 on contract from Parasound we measured every finished unit at 50cm distance and a point directly inbetween the woofer and tweeter. Response was very flat there, but no where else. And this was a time-aligned (sloped baffle) speaker with high order filters that actually sounded very good, in my opinion, if your ears were in exactly the right spot.
So now we arrive at the age of constant directivity speakers and digital micro-equalization, and mic measurements have become pretty much completely irrelevant. Strong words, I know, but the fact is with very precise and flexibile EQ integral to speaker performance, any bad measurements that depend on mic position, measurement protocols, or environment (anechoic, semi-anechoic, reverberant, in room, outdoors etc) are correctable. If Stereophile shows a 5dB peak or trough in their measurements you can tell the measurer to go into processor program and remove the deviation with a few clicks. Change the test parameters, change the program. Potentially any measurement made on the speaker under any test conditions can come out near-perfect.
It looks like designers will have to focus, in future, on improving directivity (constant directivity being "best" of course), lowering distortion (I have a way of doing that, but no one has to do what I do), improving transducer linearity (so the computer has less to do, since running out of computing power is the only restriction on your ability to correct speaker performance), reducing diaphragm noise (the materials I like are the least noisy, all other factors being equal), and other matters I won't discuss, not wanting to give the competition a leg up!
In other words there is still plenty for the designer to do, but those who believe "flat at all costs" is the primary measure of speaker quality are naive, and will eventually find their work noncompetitive in the marketplace.
