In my experience, there is no such thing as a 'reference' when the results vary by up to 20dB from manufactor to manufactorer's auto EQ attempts within this hobby.
Such a range, (20dB) pretty much entails the whole range of subjective listening volumes defined in this poll.
The following chart plots an omnimic sine sweep played at -12dB on eight different receivers in my room, after each had performed it's auto room EQ calibration with identical speaker placement for all runs, and auto EQ mic placement performed on a set of standardized micstands, with the exact same starting positions for each processor (as much as possible - as defined in the test thread). The omnimic was recording the FR plots near the auto EQ starting position capture point (position 1). You can see the omnimic is a gray mic in the center of the picture.
Who would have expected this much variance between calibrated reference in these enthusiast products?
(AccuEQ, ARC, Audyssey, Dirac, MCACC, Trinnov, YPAO)
Test thread on autoroom EQ comparison - for those who haven't seen it.
And if that isn't disheartening enough --- how about this variance I found between the exact same processor = an Onkyo PR
-SC5508 and a couple different stock Audyssey multi-EQ XT32 mics with an identical part number on a test run a different day. So even on a single processor, depending on the luck of the draw with the Audyssey mic lottery - you could have up to a 5dB variance -- this confirmed with my small sample size of only 2 mics tested. The only difference in this test was that I used two different Audyssey multi-EQ XT32 microphones. One multiEQ XT32 mic from a Denon 4520CI, and the other multiEQ XT32 mic from the Onkyo PR
-SC5508 --- both flagship receivers, both mics have the SAME part number. The test was performed back to back with no change in speaker, or mic positions....just swapped out the mic and ran the same Audyssey calibration again and measured with omnimic after both calibrations. One has to assume that while the biggest variation is at 7kHz and up in my single test, so basically double the treble ---- that the difference could very easily be any part
of the frequncy response variance mic to mic. So one person's subwoofer might potentially be producing nearly double the bass volume (6dB is double the volume) in one room vs. another with the entirely feasible 5dB variance I noted in a couple of stock XT32 mics ---- and that on the SAME model processor with the supposed same room auto EQ system.
Here is the thread link to that test: