Can't really comment on how Dirac or GLM decides what to fix or not.
With my manual EQ... Depends in how narrow those peaks are and where they are located. EQ done for one space will affect the response at other adjacent areas... I could cut some of the left-over peaks post my main listening position EQ, for example -- but it has to be done carefully.
Spatial averaging is
always better... Below I used 54 sweeps. 9 sweeps per channel at each sitting position, and using 90 degrees calibration.
View attachment 114329
I suppose if I did some really "strategic" PEQ in the bass it would like the ff.
View attachment 114335
Predicted curves post EQ
Notably, I can sacrifice a little bit of SPL between 200-290Hz in the main listening position for a better response in the other seats, I guess.
There's also a huge null around +130Hz in both channels in my MLP which I've filled up BUT:
View attachment 114331
Wavelet spectrogram 1/6 res, 25dB scale
My boost around that null/node doesn't actually fill itself completely -- it's still a black hole -- but that extra energy additionally expands the decay in time which does fill-in the hole in the simple SPL magnitude graphs.
View attachment 114332
It also does not look too different from an unwanted resonance/extended decay... which could potentially sound bad afterwards, or not -- in this particular case, it sounds totally fine.