I see this is an old thread, but here's how I worked it out on my Panasonic VT50. (I should mention I'm not an ISF, just a dude with a meter)...
I don't know to what it extent it's just the CMS on the VT50 but I found that the lums controlled the linearity of the saturation levels for a color.
I selected a reference color, red, since I think that's the most important one (because of flesh tones). I went back and forth and adjusted its lum setting and the main color control until its saturation level was accurate at 25%, 50%, 75% and 100%.
Guess what? That was the correct lum setting as measured normally, and it left the main color setting at 50. Perfect!
I never touched red again.
For the rest of the colors I first calibrated the lums in the usual way since they all need to match and I was not willing to mismatch them to ensure linear saturation for all the colors - (though sometimes I wonder how that would look).
Then I adjusted the discrete sat levels. Without using the lums to try to get them linear, I didn't know what to do but choose what sat level I cared about the most. I went with 75%. I think 100% is standard for Rec 709 but IMO by the time a color gets that pure it's kind of a "special effect" anyway. The only exception was yellow which I ensured was exactly correct at 50% because it's critical to flesh tones at the lower saturation levels (or that's my supposition anyway).
This required another adjustment to the lums, and so forth - one of those endless calibration loops you have to decide where to end.
Anyway, most of the sat levels worked out close to linear when the lums were correct.
I think it looks great. Sometimes the saturation seems high, but I've found I usually think correctly calibated TV's have too much color, so I'm not sure. When I have time - just for kicks - I'm going to experiment with setting the lums purely so that the saturation levels are linear, (if possible), and see what that looks like. It would be close settings to the ones selected from light output relative to white, and possibly less vulnerable to ABL(?)