In my practical usage, I will state again, that despite what the above tests might or might not show, the color saturation and variation I achieved with the older DP2, was quite usable, and did deliver a wide color palette. And when boosting color and vibrance sliders in post, there was no apparent color noise, and the variety of color did not diminish...at least with shots done at ISO 100 or ISO 50. This is not true of all the bayer-sensored cameras whose RAW files I have edited, including the 5D3 and my 6D. Of course overall, the file is better from those two than from the DP2 (not even discussing the file's resolution dimensions here). Their sensors are huge by comparison, and their implementation is far more developed and evolved.
But it's really just something you have to be open minded and experience for yourself. You can't judge Sigma's sensor alone, based on tests and charts. I suspect the Merrill sensor is significantly better, as well.
Basically what I am saying, is that when editing RAW files, the primary colors, become blown out very quickly, when you apply boosted saturation in post editing of the files in Lightroom or ACR...with bayer sensors...where they do not with the foveon sensor I used (again at low ISO...at higher ISO, yes it was not usable for color...but no one has said it is).
You can attempt to discredit what I just said, or say it's not a valid point. I don't care. It's still the reality of using the foveon sensor. Sure, you could argue that such boosted color saturation is not a realistic interpretation of the reality the camera was capturing...but I could argue that a big part of what it captured IS REALITY, it is just representing it in a way that flies in the face of the philosophies you adhere to, and thus you close your mind to other interpretations that exist in reality, whether you like it or not.
Again, thus far, the only digital camera sensor that has achieved an interpretation of "true color" in the lateral plane, is the foveon. Film did it in the lateral plane, and essentially in the longitudinal plane (since the emulsion layer was very thin). Bayer sensors certainly rely on math to determine color...to argue the opposite is to lie and ignore the obvious. So an argument based on the supposed "flawed math" needed to produce an image with a foveon sensor, is a flawed argument itself.