I did say that the print on ID Gloss looks exactly like the soft proof which in turn looks like the original. I think I explained the conditions too, but I haven't tried on matte paper, which I also said.
I've reread your posts above and don't see anything about actual prints -- I'm sorry if I'm just not understanding you. I believe what you say about the prints matching, but this is not what I see with any media I try. I would really like this problem to just go away with a "oh, look, I'm doing it wrong" but I can find nothing, anywhere, in my configuration that indicates that I am doing so. Remember that I have tried on the Mac and on XP (two different computers) and the failure mode is identical each time. I have pounds of printout that all look wrong in exactly the same way, except when they are even more wrong because I have intentionally double-profiled.
From my analysis of the problem, I am wondering if the driver or the printer is ignoring the "application managed" option, when it is selected, and performing some sort of colourspace conversion or automatic print adjustment, so that when the raw RGB comes from the application it is translated even though it is told not to.
I wonder further if my printer, having had multiple firmware upgrades applied, somehow works differently than printers that have more recently shipped with a virgin firmware installation.
I wonder if it is possible to install a "null" profile into the printer for a paper type, such that even if it was told to manage colour itself, it would actually not perform any lookup. And, thus, it would be impossible for the printer or driver to screw up application managed colour at all.
Does anyone know of a pre-built "null transform" ICM or ICC profile I could try this theory with? I could probably convince argyll to create one, but if one already exists I would be more sure of doing it right and not spoiling the test.
Really, all I want is my printer back. I'm so weary and fed up.