A ccss with RGB only (I created it from RGBW measurement and deleted the last entry) will make dispcal to crash. A ccss with RGBW is closer to the ccmx from RGB. So, to sum it up: The matrix from RGB gives the best, the matrix from RGBW gives the second closest (but similar) and the ccss (doesn't really matter how many but must be at least RGBW) the worst match to the spectro. (It doesn't mean it's the best match to reality, just to this spectro but this is the best I have and why would I trust in the colorimeter better...?) A matrix from RGB seems to give a nice match to the spectro even if the display isn't additive. (But it can be tricky if not practically impossible to work around the faulty DSP and measure the native RGB primaries...)