I agree with you, but would like to see measurements of the magnitude of the effect from the second pair. The first several speaker comparisons I did, I swapped locations of the two pairs and established (to my satisfaction) that the difference associated with location (when side-by-side) generally was small compared to the difference between speakers.
However, as I concluded when comparing amps, the difference associated with location exceeded the difference between amps so as to render my attempt at a comparison under those conditions as worthless (other than to conclude that the difference between amps was slight enough as to generally look at changes in other parameters (like speaker location) instead of amps)!
Measuring the effect is easy- test with them in place and then do it without.
If you have REW, run it with another speaker in the room without connecting it to any equipment, then shunt the terminals on the unused speaker. The response will change because shunting the voice coil prevents its movement in the magnetic field (you can test this by pressing gently on the cone and noticing the difference in how easily it moves with/without the shunt). Large drivers are often shipped with a shunt wire because of the extra weight of larger cones for this reason- if the box is dropped, it may not show any damage because of the packing, but that doesn't mean the voice coil won't bottom out (assuming the magnet doesn't have the bump that allows full excursion, to eliminate this).
Shunting obviously won't do anything for diffraction, but it does show the reasoning behind people recommending that only one pair of speakers should be in a demo room.
