It seems like it was a good a idea to create a ccmx from R,G,B measurements only (no white, no multidim steps, just three signle channel colors with both sensors and that's all). Even though I didn't measure white, the colorimeter agrees with the spectro about the white point better than ever. The difference is no higher than the usual deviation between the repeated measurements with the individual sensors. And it keeps reporting me the same white temperature as the spectro even after I set the white point to a different temperature! I created the ccmx with 6500K white point settings, changed the device white point to 5000K and the colorimeter read 5000K (withing the repeateability range but very-very close...). That's what I expected from a colorimeter profile! For some reason, ccmxmake doesn't let me create a ccss file without white measurement.