Dean,
I would like some clarification. Are you saying that detail is lost in out of gamut areas or merely color in the detail? Somehow, I think it is the latter, because,you can completely desaturate an image without losing detail.
Rob
I'm saying detail is often, but not always, lost in an out of gamut area. It depends on exactly what makes a detail visible in the first place. For instance, two adjacent pixels of the same hue and saturation can be differentiated by luminosity. If both are out of gamut due to excess luminosity then they become indistinguishable. Two pixels can also be distinguished by a difference in saturation. If both are out of gamut due to being over saturated, the pixels become indistinguishable. In both cases detail is obliterated.
Gamut is actually a volume with hue (color), saturation, and luminosity coordinates (HSL), though there are other ways to describe it. The the outer surface of that volume is a set of HSL coordinates beyond which you go outside the available gamut. If there is a gradient in hue that's all of the same luminosity, and the whole gradient is out of gamut, you'd get banding of the hues. Desaturating won't help because it's all the same luminosity and there would be no differentiation. I doubt you'd want to change the hues (colors), so decreasing luminosity is the only option left to bring everything back into the available gamut. If it's the same hue and all out of gamut there's a potential option of decreasing saturation or luminosity or both to get back the gradient. In a nutshell there are three things you can change to get an area back within gamut, and one is hue or color, which you probably don't want to change, so you're left with saturation and brightness (luminosity).
Suppose there's a gradient of the same hue that varies in luminosity, as sometimes happens within a cloud structure at sunset. Part of that gradient is just on the edge of the available gamut, and as the luminosity increases over a distance it gets further and further out of gamut. That entire area of the print will show no gradient. Instead it's a just blob of that single hue at the maximum luminosity available for that hue in the gamut. In this case if you decrease the luminosity so the part that's furthest out of gamut is back in gamut, you'll recover the whole gradient. But, since H, S, and L all interact, it's also possible you could decrease the saturation to get the gradient back. What works best really depends on which of those 3 coordinates (really 2 because we don't want to change color) is most out of kilter.
I'm sure I left something out or screwed up some detail, but I hope that's adequate to convey the basic idea.
