So if you dont like the clipping, you'll accept the Perceptual one?
In theory that maybe right. Here is an example of an image (tight crop). The Relative is very ugly because the shirt is totally ruined. The Perceptual is washed out (hopefully you can see it that from the upper corners).
Exactly, this is a good example, and not that uncommon for saturated colors. Letting the profile conversion do its thing will result in either featureless color content when Relative Colorimetric rendering intent is used, or an overall reduction of saturation, also in areas that were originally not in need of correction.
The problem is with the way one wants to, or is able, to address the issue with a bit more intelligence. Only selecting the color tones that will clip, may cause a disconnect with the immediate neighbors of a similar color tone that do not clip. So one typically creates a correction that has some sort of feathering applied for a more gradual transition into non-affected areas. But to do that, a selection of the problem areas is required.
For me choosing the shirt and tweaking with saturation, hue and lightness in an adjustment layer plus using the Relative gave the best result.
Which brings us back to the original question, of how to select the OOG colors in Photoshop. I suggested one possible method of creating a selection, not only of the pixels that are OOG, but also their OOG magnitude. Relatively small amounts of isolated clipping require less attention than the more severe ones that will cause visual loss of detail.
Therefore it would be nice to be able to select the out-of-gamut pixels properly.
The method I've suggested earlier (and converting the created Hue difference layer to a selection with a bit of feathering) will allow to do that, but perhaps Andrew or someone else has a better solution (other than not using Photoshop).