We do now actually understand what is happening, and what, um, err, well, 'stuff' is done to try to mitigate the issue with other integrations...
While the issue is often seen in low-light readings, it can actually happen in any highly saturated patches, especially red.
What we are going to do with ColourSpace integration is still debatable, as ColourSpace actually generates better, more accurate LUTs when the Rev-B and later probes provide the correct, zero value, readings for such problematic patches, due to the advanced error handling of the Colour Engine in ColourSpace.
But, when performing manual calibration that is not really what is wanted...
So, we are likely to add our own 'workaround', that hopefully will be better than the alternative options.

Steve