Yes, the explanation was rather convincing. From what I understood, it works this way:
- imagine that you are exporting to disk to sRGB (like many do)
- colors that easily fit into sRGB (central parts of the gamut) are left untouched
- colors that are near the borders of the sRGB gamut are only slightly shifted
- colors that are outside of the sRGB gamut are shifted to the borders of the sRGB gamut so they fit
The amount of shifting to fit follows a non-linear rule so that hard posterization/blocky areas of uniform color are avoided. There’s some sort of “knee” to avoid simply truncating colors outside of the gamut.
Perceptual would shift ALL colors. Colorimetric relative would do a hard truncation of the colors outside, leaving untouched those inside (but leading to loss of detail in heavily saturated areas). DxO’s method tries to match the best of the two intents, and the aggressiveness of what happens at the borders of the smaller color space is dictated by the “protect saturated colors” slider.
Correct me if I had understood wrong, but it seems to me that it’s working well.