I'm going to take a stab at this and guess that the explanation means that, should pixels in an area clip to white in only 1 or 2 channels out of three, then the additive mixture of the R, G and B components, including the clipped components, may exhibit artificial color fringes because the clipping to white is not present in all three channels. This could be due to overexposure and white balance changing the relationship between the clipping in the R, G and B channels.
I mocked up an artificial example in PS by making a gradient from black to white in each of the document's R, G and B channels (the channels in the attached screenshot "clip" at pure white). The example simulates, perhaps, a high contrast edge with some soft but narrow falloff of tone from shadow (i.e., intact pixel information) to blown out white (a specular reflection, for example) from right to left across the image area. I offset the green gradient slightly, so that there is an area in the image where the red and blue channels blow out while the green channel still has some pixel data remaining in an area of pixels that overlap the red and blue blown pixels. The result in this region is, as you might expect, a magenta fringe (100% blue + 100% red relative to not 100% green is approximately magenta for purposes of our example).
Maybe this is what the Sigma person is saying? In areas where there are narrow regions that contain pixels that have one (or two) channels blown but two (or one) channels intact, color shifts can occur - probably more so in regions where the transition occurs within very few pixels, making the "edge" between intact and blown pixels more apparent and a fringe more visible.
Some raw converters attempt to reconstruct these blown out areas with luminance or color propagation highlight reconstruction algorithms, using data from the intact channel(s) in the pixels in these regions to infer the color data in the blown channels in these pixels. Perhaps the Sigma conversion algorithms do not perform this? What little I know about Sigma sensors is that they are funky in how they capture light and how the raw files need to be processed. If it is something specific to Sigma, then my guess may even be more off base than it already may be.
kirk
