I was experimenting with a method of creating dual quad ink sets to determine if it was important that the two profiles be as close to each other as possible (to me it seems like it is). Depending on the inks used as colorants, they might have a different L* reading for the same ink limit for the same shade from one set to the other. An example: Carbon shade 2 might be 1.3 at 40% and the Selenium might be 1.3 at 45%.
With that in mind, I went to make a workflow to easily ensure that they are the same (or as similar as possible) from set to set. This means matching the profiles for each color set as closely as possible: the max density (or ideal ink limit) for each shade; the exact overlaps for the same shade from one set to the other; the total density; and the final linearization. Its seams like it does make a difference when examining smooth bullseye gradients, but that was with just one day of testing and it might take some more experimentation.
I will see if I can easily post my Excel worksheet for finding the exact cross-over points. That alone might solve some these problems.
Until then you might try the following to see if it helps with linearization, and creating smoother initial profiles.
You can use a GRAY_OVERLAP of 20-40, and then a GRAY_GAMMA of 1.2-1.4 to offset the increased density of bumping up the overlap. This should help in smoothing out the bands where the inks are overlapping (and help when blending two final sets of curves)
If you are still having problems with the 6 partition profiles, try increasing the GRAY_OVERLAP to 60 to about 75, and the GRAY_GAMMA to 1.6-1.8. This will mean you might need to (read: should) lower your ink limits and/or make the max density of each shade be evenly spaced, rather than setting the ink limits globally. If the dmax of shade 1 is 1.65, then shade 2 should be between 1.35-1.4; shade 3 between 1.1-1.2; shade 4 between 0.8-0.9; shade 5 between 0.5-0.6; and shade 6 between 2.5-3 (using density to establish ink limits is easier to see than using the L* readings—I've included the formula below—you first need to convert LAB_L to XYZ_Y, and then do the density conversion)
L* to Density Conversion Function
If L* > 8
Y = ((L + 16)/116)^3
Else
Y = (L * 27)/24389
Density = -log10(Y)
Hope that doesn't further confuse the matter,
Richard Boutwell