I have noted something of interesting about CS generated cube patchset with extended scale range.
What in principle happens is that a 235-cube level is always inserted. The number of patchsets is kept the same by decreasing by one the number of levels between 16 and 235. I have written in principle, as, the algorithm is much smarter, as when the density of level allows for that, add more levels between 235 and 255.
So, if you start a 3-depth cube you have combination of RGB 16,235,255 (0%,100%,109%) vs. the legal case, where you have 16,118,235 (0%,47%,100%). In the same way, for 4-depth, you have 16,118,235,255 (0%,47%,100%,109%) vs. 16,81,156,235 (0%,30%,64%,100%) and so on.
The outcome is that the sampling of the same measured gamut is different depending a legal/full or an extended patch scale resolution is used. Specifically, the 100%-109% area it always oversampled, at the expense of 0%-100% one.
Does that (i.e., 235 level inserting), make sense as:
We are removing the need for a 235 (8 bit) 100% patch when using Extended Patch Scale.