For Nikons, most of what you've observed has to do with the high IR sensitivity of a D100. Back when I shot a pair of D100, I used to use IR blocking filters (Heliopan or B+W) to keep IR contamination from messing up skin tones, plant greens, browns, and yellows, and any dark clothing. D1, D1X, D1H, D2H, D70, and D50 all shared this problem to one extent or other. D2X was the first to incorporate a dichroic IR blocker, followed by D200, D80, and D40X. These cameras lead the industry in color accuracy.
As to the rest, sorry to say that it's mostly blowing smoke. I've put cameras on a monochromator (a complex device used as part of the process of profiling cameras for writing good raw converters) and ran spectral sensitivity curves of the sensors. When you're talking about what camera can feed the best raw data into a conversion process, Canon is actually behind Fuji and Nikon (the same Sony sensors used by Nikon are also used by Pentax and Sony). So, in a profiled workflow, Canons are more difficult to profile, and will have greater problems with observer metamerism once profiled. This would technically put Canon at the bottom of the pile, just ahead of Sigma (the Foveon chip mentioned elsewhere in this thread), for color.
When you're talking about onboard conversion, Canon is the equivalent of Velvia, pleasant colors, but totally inaccurate. Fuji onboard conversion is a compromise between accuracy and "pleasantness" that many shooters find useful in a "shoot to print" workflow. Nikon is the most accurate, which paradoxically, makes the pictures the most unpleasant.