D driver, it is my understanding that both effects contribute. Increased aperture even helps with CVOR site errors, there were papers written on this in the '40s, but as antenna spacings approach 0.5 wavelength bearing ambiguities appear and the result is not usable. So, DVOR has an inherent ability to discriminate against weaker reflected signals, plus it removes a constraint that lets you really go to town on the array size.
I was hoping to find some quantitative comparison of these effects but have not had any luck. This research report
"Effects of scattering by obstacles in the field of VOR/DVOR" looked promising, but despite its title it doesn't actually contain any DVOR results.