For what it's worth, I get the best scans with matte paper with profiles made from prints on the same matte paper. With glossy paper I get good colorimetric scans for higher luminosity patches but the darker ones not so much. And that's after correcting for wide area crosstalk I have discussed and provided techniques for correcting. I have some evidence it is due to stray light reflectance that has a larger effect on darker patches. This is a bigger problem on glossy than matte.
I have been thinking of altering my targets so as to group low L* patches together. I think this may result in better profile performance and low L* but it's just speculation until I get around to testing it.