They may be choosing from a selection of parts that are available. It's also possible to process out smaller smaller than it is actually captured. You might end up with better detail in saturated primaries, given the way they're distributed. In a square shape you would have 2 green pixels, 1 red, and one blue. It's not as bad as it used to be, but the bayer array never seemed terribly ideal.
A number of years ago I worked with the first 2 generations of the 1Ds as well as the Phase One P30 and P45s. Megapixels can make a difference in detail resolved. 8 to 12 isn't a very big difference at all. Some of the 645 sensors with no anti aliasing filters resolve considerably more detail even compared to dslrs with similar pixel counts, although in most cases the dslrs hit the point of good enough several years ago. Even the software that does the debayering and rasterization is a huge factor. You can get a much more pleasing result with some of the modern software revisions today (Lightroom, Phocus, Capture One, etc) even on older dslr cameras. What's interesting is that the tiny cameras that go into phones have certain design limitations yet lack others. There's no mirror so achieving wide angle views doesn't require a retrofocus (reverse mounted telephoto) lens design.
"Some of the 645 sensors with no anti aliasing filters resolve considerably more detail even compared to dslrs with similar pixel counts"
You just changed your story, and correctly that pixel count is not the factor but the processing and sensor quality, as well I would argue, the glass used.