An observation:
I printed the Gamut Test File courtesy of Andrew Rodney while comparing prints with sRGB and ProPhoto profiles
Something I didn't notice before:
The file has at the top six color patches.
At the bottom there are fourteen balls.
The blue patch AND the 14th (blue ball) printed:
On Epson Premium Semigloss
Epson Semigloss with SPR2880 Epson Premium Semigloss Prophoto Perceptual ....BLACK
Epson Semigloss with SPR2880 Epson Premium Semigloss Prophoto Relative ...VERY DARK BLUE (Almost black)
Epson Semigloss with SPR2880 Epson Premium Semigloss sRGB Relative ... BLUE (very much more like the image in the screen)
I made a comparison with Harman Gloss Baryta
HAR_Eps2880_PK_GlossBaryta Prophoto Relative ... Blue (similar to that on the Screen)
The blue-> black looks like too much
Any comments?
This, blue ball printing black, is a well known issue with many canned Epson profiles.
It is not a technical profile defect. Almost all the colors in the "blue ball" are imaginary in that they are outside of the CIE XY gamut. Further, they actually have a L* value below 1. In a sense they should be an extremely dark if not black, blue. You can check this by loading the image in Photoshop, converting to Lab colorspace, and examining the blue ball with the Info tab. Note the extremely low L*, which corresponds to Luminance.
The reason that the ball doesn't appear black is because the conversion process to your monitor's colorspace raises the L* value. A lot. You can see how much by converting to your monitor's colorspace or, sRGB, Adobe RGB, or most any other standard RGB colorspace.
However, the conversion to the printer device space goes through a completely different path. The Lab values are used to look up, in 3D LUTs, the printer's RGB values. Since the colors, being imaginary, are well outside the printer gamut how these are determined depends on the algorithms used to create the profile. There is no specification as to how OOG colors are mapped. If you compare this against a profile created by XRite, you will see the XRite printer profiles will raise the L* value much like conversion to standard RGB spaces.
This has nothing to do with profile "accuracy." Both may print in-gamut colors accurately. And both should show softproofs that model what actually gets printed. If you look at the Epson softproof you will see the ball also comes out a dark blue/black while an XRite profile softproof is a much more pleasing blue.
However, these Epson profiles are much more problematic when printing from ProPhoto when some of the colors are well outside the printer's gamut. This is most likely to occur when pushing saturation working in ProPhoto and not viewing the soft proof result. Especially with these Epson profiles.
If you first convert the image to Adobe RGB or sRGB the L* will be raised and the blue colors will be turned into "real" colors inside the CIE XY gamut. Do this and examine the Lab values using the Info tab. Since these colors are much closer to the printer's gamut, printing them will produce a reasonable blue ball with both the Epson profiles and XRite (or most other) profiles.
I consider this image a good "gamut stress test" as it shows how smoothly the profile renders OOG colors. I like to use this, in combination with something like the Kodak DPI Colordisc image which is largely in gamut. That one should print good images with either profile.