The OP headline doesn’t gel all that well with the described findings, it seems. Too many factors are at play determining "bigger soundstage" than mere driver size alone, though better "room fill" is more in line with the use of larger drivers/overall, summed radiation area. Larger radiation area to me generally is about better scaling (which, admittedly, is an aspect of soundstage presentation) and an added sensation of density, impact and ease.
As pointed out by Duke larger drivers have a narrower dispersion pattern, and especially when coupled to horns the sonic marker of this is intensified. As to multiple smaller drivers vs. a single larger one (per channel) a line array of smaller woofers/mids would act as such and limit vertical dispersion, whereas a single, larger driver acts as a point source with narrow dispersion overall. I’ve never much cared for the sound of the multiple smaller driver configuration vs. a single larger one, but whether that comes down to the dispersive nature mostly or other I couldn’t say (I do prefer what emulates a point source presentation over a line source ditto).
Typically the use of smaller drivers involve low sensitivity designs, where the larger driver designs I'm thinking of use high sensitivity units - this is not insignificant either; used within their upper band confines, not least when high-passed down low (which may be a given in the first place as such drivers are usually LF-limited, and thus in the need of subs augmentation), a high eff. large driver sports very low distortion and ample headroom with limited cone movement and inertia build-up. Smaller, low eff. drivers, even in multiples, simply can’t replicate these traits, though they’ll extend higher. Personally though I’d go for a lower cross-over in the ~800-1.2kHz region (with 12-15" units) and have one large driver (or the same in duals) cover the "power region" down ~100Hz, give or take, and avoid a x-over in the upper midrange typically found with smaller drivers. One could also use smaller drivers (8-10") in the same frequency span - say, in a D’Appolito config. - and have very respectable results as well, such as found with the (sadly now defunct) S.P. Technology speakers (by Bob Smith) in conjunction with a waveguide.
