Hayo,
We are now using LittleCMS for cross-platform color management instead of Apple's ColorSync (they deprecated almost all of the methods we were using and replaced them with new ones that are poorly documented) but I expect the color to be nearly identical.
Yeah, quite annoying that Apple deprecated the methods and then chose to not properly document the new ones 
It may be that setting a color to pure red or pure blue gets wrapped to black on your profile with LittleCMS. If I were to make it just one less (254 vs 255) for the red/blue component that it would work fine in your case, Hayo.
Hmm, I guess this might help but I think it does indicate a bug in the LittleCMS. What if the original image already contains such a colour (very well possible, especially with e.g. artwork)? Does that also not get rendered incorrectly?
I don't know. Try creating a TIFF file with half of the image pure Red and the other half pure Blue and then view it on the affected system with PM6.
Hi Kirk, I reinstalled my old laptop and added a demo version of PM6 onto it to try and replicate things. I created an image file with all primary as well as all mix colors with their levels set to 255. I then assigned it the colour profile of the LCD Display (so the colours would come out unaltered, or at least so I had hoped). Viewing this image did reproduce the highlight problem for the white square that was in the image, but all other colours showed normal. I then measured the colours (using the digital colour meter application) and found that pure red, instead of having 255 as value, “only” has 254, the same for pure blue, pure green became 0,255,1 and magenta 255,0,254. Cyan and yellow on the other hand where fine (0,255,255 and 255,255,0 respectively).
I remember discussing these little discrepancies (rounding errors?) before, but I think they are exactly what is causing the problem in some of the cases. I would be happy to try a version where you set the highlight colours to 254,0,0 and 0,0,254, to see if that solves the highlight problem. But to be honest, I still think we then still might have an issue with some of the colours inside images i.e., ones that map onto 255,0,0 and 0,0,255, and perhaps even 0,255,0 after profile translations. Granted, I have not been able to produce colours from an image myself yet, but this is likely possible.
P.S. I do find it suspicious that if I have a file with the colour profile of the display, I do not get the exact numbers for the colours as I had in-file; feels like a translation error somewhere in the pipeline, but at least the values measured are the same for the file viewed in Preview, so if it's an error, it's outside PM
