OK, a little about measurement systems;
1.) that ruler-flat line that everyone strives for- sorry, it doesn't exist. What you want is a smooth transition from octave-to-octave, frequency-to-frequency. If you look closely at the FR traces (spectrum) you will notice that the horizontal lines are 1db increments. Inspecting the midbass range, you'll see that there are adjustments of roughly 8 db (that's substantially less energy in that frequency range). Again, you can't look at the trace and say "oh my, look at the swings from freq to freq" because that will always be there- you cannot eliminate them. You look at a trace and interpret what the trends are: midbass area is bloated by 6-7db, upper-treble (10k+) is depressed by 4-5 db, then you try to solve for those with as few, wideband/minimum phase filters as possible.
2.) In the first screenshot I posted, the middle section of the screen shows phase response (referenced to the actual source signal) measured by the microphone. You'll notice that from 31-16khz the phase response POST-Dirac remains within +/- 150 degrees and stays pretty smooth, but the phase response without Dirac is greater than +/- 180 degrees; in some cases 360 degrees (delayed)- and that shows that the Dirac processing is in fact reducing the group delay by roughly 25%. I posted earlier that frequency response didn't tell the whole story, here is another significant part of the story.
3.) The result is that Dirac actually makes these 2 speakers sound better than they have any right to in this particular setup- being 5 feet apart setting on a table. The result of the Dirac optimization was that the image became much more 3 dimensional, the bass became tighter (expected, since the midbass bloat was gone), and there was greater resolution brought out of the recordings (upper frequencies being compensated).
The FR traces that I posted were smoothed to 1/12 octave from the normal 1/48 octave I measure at- in order to show the transitions a little better. Essentially, what happens at 1/48 is that the trace from roughly 1k up looks like a richter scale trace, and you're doing a visual average- or you tell the software to average the instantaneous readings (what I do typically- either 1 second averages, or choose the slow response average setting).
I've never used Audessey, I cannot comment on it. Generally, I use commercial DSPs or loudspeaker management boxes and measure like I have here.
Please realize that I didn't post everything I measured- this is not the thick of this experiment, not by a long shot. What I've posted today is the beginning- a glimpse of what it is that Dirac does- for real, not smoothed to 1/3 or 1/2 octave. Our brains do a tremendous amount of averaging when we listen- the microphone and software show us what is really being played, our brains tell us the story.
In any event, the rig I used today was as follows (and will remain the same throughout the experiment) Macbook Pro (2011 model w/ dual core i5 processor), Presonus Firebox (Firewire 800 connection), Earthworks M23 Microphone, Manfrotto mic stand, various 1/4", XLR, combo cables, SMAART v7.(x) - this is in flux, I may upgrade to the latest version (7.3 or 7.4) before the Trinnov shows up. This is a similar rig to what is used to certify live venues, auditoriums, theaters, etc. It's what is used by most professionals and accepted as a reliable measurement system by pretty much all acoustic consultants.
Thanks for reading, I hope everyone else finds this as interesting as I do.