I think I said ABW “dark mode” was as linear as one might expect.
In my testing context I am sending a wedge with L* patches, directly to the ABW driver via Photoshop, then reading the patches with an Eye One. Of the 5 modes available (light, normal, dark, darker, darkest) I found “dark” was the most linear and, (factoring in the realignment of grey values due to the differences in DMAX,) acceptably linear to build a profile on.
Without establishing parameters we may be all saying the same thing with different tolerances for variation. It goes without saying that no matt paper will ever be linear, unless we are not using the definition in the same way. If linear means all the input L* produce their real world coefficients in ink density that would be impossible. A paper with a DMAX of L*22 will never be able to redistribute the missing 20L* and regain perfect alignment. You can’t make 100 input grey values be the same as 78 output grey values.
I don’t think perfect alignment is the goal, intelligent realignment is.
Eugene