MORE DATA
Signal..SPL
..NONE...40 to 55 background noise
..-80....42
..-75....46
..-70....50
..-65....53
..-60....58
..-55....63
..-50....68
..-45....73
..-40....78
..-35....83
..-30....88
..-25....93
..-20....98
..-15....103
..-10....108
The signal was white noise, as before, but limited to frequencies below 400 Hz, and reproduced by my subwoofer system. (Easy to do. I just muted the HF and adjusted the X/O frequency up to 400 Hz). From prior experience I know that the warble tone would have rattled things. The mic was positioned about one foot from the 15 inch driver.
This time I set the signal level using the preamp volume control, and read the resulting SPL. When reading the SPL for the range below about 60 dB I took the lowest of rms indications over about 30 seconds, which corresponded to a lull in the traffic (background noise). For the higher SPL readings I took the average, as before.
Plot the data and you will see that there is almost no suggestion of decreased sensitivity at low SPL, which I attribute to the greater care that I took to minimize error due to background noise. And anyway, the SPL range where the data is not perfectly linear (for whatever reason) is so low as to be almost inaudible, so it wouldn't matter anyway. I am particularly happy to see no compression for high SPL, which would be a worse problem.
All of this is for my speakers :-). Maybe yours are different :-(