Btw, Graeme, I've seen, that you have implemented, but not activated, CIECAM clipping in profile. Is there any reason for not activating it?
No, it's meant to be on. An oversight (I think I turned it off while investigating some performance issues a while back, and didn't notice I'd left it off), but it doesn't do much, unless you are far from the gamut hull.
I had inspected profiles generated by PP, and they seem perform a lightness compression, even for the colorimetric intents, which may be pleasing, but which is acatually not correct and does not comply with the ICC spec, IMO. The colorimetric intents are expected to reproduce all in-gamut colors colorimetrically correct, and to clip only out-of-gamut colors. If you look at profiles generated by other profiles (e.g. ProfileMaker), you'll see that they behave similar as Argyll for colorimetic intent. However, the ICC spec leaves open, _how_ out-of-gamut colors should be clipped, so this is implementation dependent. For a pleasing reproduction with gamut compression, actually perceptual intent is intended.
Colorimetric intent simply clips the colors to the closest point on the gamut hull. Closest as measured in the space the gamut mapping is represented in, which is either L*a*b* or CIECAM02 space.
Absolute colorimetric is generally use for hard proofing, where the proof is intended to lie next to the original, and match it as closely as possible. For this to work well, the instrumentation needs to be as accurate as possible, e.g. preferable the same spectrometer used to measure the sample patches of both the target (input) medium, as well as the proof (output) medium. Off hand, I would imagine the limitations of using a scanner as a colorimeter would show up rather badly under such exacting circumstances.