Took some measurements just now.
No smoothing at all on this first set.
All Channel Stereo - Track 2 off the Omnimic CD (which is left and right speakers together, and then upmixed by the Denon x7200wa to all channels) - No EQ, No Audyssey.
Same thing - but just standard Stereo on Denon - Track 6 Left and Track 12 Right off the Ominimic CD - No EQ, No Audyssey.
Retook the same scenario measurements with 1/6 smoothing (which I read is about what we humans can actually hear - some even say more like 1/3 which is smoother still)
All Channel - track 2 - still no EQ/Audyssey - 1/6 smoothing
L and R Channel - track 6 and 12 - no EQ/Audyssey - 1/6 smoothing
Yes, I run my bass hot, (and subwoofers were on) but surely not as hot as
, per your question earlier, when you said you get more bass in all channel stereo - I think that's only if you use Dynamic EQ. I've been avoiding that lately. When I turn to all channel stereo without EQ, it sounds like less bass, which is also what the graph shows at least bass in relation to mids. (the overall bass SPL levels are higher with all channel stereo, but it doesn't sound that way proportionally)
Consider left/right graph, bass is about 15 dB hot. On All channel stereo, it's still hot relative the mainspakers, but quite a bit less relative the mid frequencies.
Oh - and I guess I shouldn't have said NO EQ on all the graphs above, because I do have a tone control applied since Audyssey is off that I set to + 2dB bass and + 1dB treble, and those settings apply to both Stereo and All Channel Stereo. But there is no individual EQ applied to any of the speakers, and no Audyssey.
All channel stereo's comb filtering with all 9 JBL CBT 70J-1 bed speakers playing (my all channel stereo has the 4 top speakers disabled) doesn't look worse than just a single left or right speaker speaker's frequency response capture at my MLP where the speakers distances/impulse response is calibrated to. I am using line array speakers, so my results may not match a more traditional bookshelf or tower speaker design.