Hi Jolyon and Cedric,
Just a note to consider that colour charts have strongly fluorescing panels. This means that relative to other panels, their appearance in the picture, and their reflectance measured with a spectroradiometer, will be dependent on the illuminant. The pictures below demonstrate this using a SpyderCheckr24 but the same is true for x-rite colorchecker. The first picture uses a blue light source with a yellow filter, the second a green light with a red filter (no white balance on the pictures). The panel at Col 4,Row 3 is particularly fluorescent, but others are also fluorescent to a lesser degree. I have yet to find colour standards that do not fluoresce (including Munsell chips!). This is unlikely to cause any problem for terrestrial pictures, but I can see this influencing the calibration of underwater pictures, especially below a certain depth when using natural illumination (e.g., > 10 m). A simple solution may be to first identify the fluorescent panels and exclude them from the calibration.
Thanks so much for testing this out and flagging the issue! I’ve only used UV dark-lights to test X-Rite charts using subjective human-vision (and decided nothing was detectably fluorescent), but I haven’t tested them properly. For now, I guess we should recommend the chart-based technique should only be used with natural terrestrial light, or the same light used for the photography. I measured the charts I’ve got using direct sunlight to control for any effects like this, but you’re right about highly modified natural light (such as underwater).