What about multispectral imaging in a very narrow bandwidth around those non-overlapping gamuts?
Let’s assume we stand outside the house at somewhat D50 light. Through the windows we see there is colder D65 light inside the house. Let’s further assume that the camera does not impose any limitations to capture the gamut of such scene.
Then comes the conversion of the demosaiced "Raw" RGB data to the working space or internal working space of the Raw converter. Here we have the option to control white-point-mapping by means of the given Temp. and Tint sliders. Now there are two different cases whether we click-whitebalance on a D50 white or a D65 white from the initial scene. And we further can construe different cases depending on the working spaces that are tried here. Let’s stay with regular Adobe RGB D65 vs. ProPhoto RGB D50, or linear-gamma versions thereof when we talk about the internal working space of the Raw converter.
Your assumption - if I understand it correctly – is that Adobe RGB could hold specific colors from inside the house which are out-of-gamut with ProPhoto RGB. And you may be right in an absolute colorimetric sense: White to saturated blue colors from inside the house can be out-of-ProPhoto RGB when we click-whitebalance on the D50 reference white of the scene. Same colors can be inside Adobe RGB when we click-whitebalance on the D65 white from inside the house.
But then we could as well have click-whitebalanced on the D65 white while using the ProPhoto RGB space, thus capturing all in-house colors and probably much more from outside the house.
The point here is that it does not seem to be possible to freeze the absolute colorimetric differences between Adobe RGB and ProPhoto RGB. It is becoming irrelevant.
The same click-whitebalance operation changes its meaning from somehow AbsCol (e.g. scene D65 white to Adobe RGB D65) to somehow RelCol (scene D65 white to ProPhoto D50) depending on the working space used. Subsequent conversion to the monitor white is RelCol anyway.
Peter
--