I opened your ICC profile in ColorThink and made a 3D graph of the gamut volume. There is definitely some funkiness going on in the magenta region - view it in wireframe at 5% steps. There appears to be compression of the wireframe mesh in the darker tones from purple through red and there are some distortion in the area of the darker magenta tones.
Screenshot:
http://kirkt.smugmug.com/Photography/Link-Share/i-kDhZNN3/0/O/Ep-ScreenShot.jpg
Perhaps when your image gets converted from its working color space to the printer profile space, the relative colorimetric intent, bringing out-of-gamut values to the surface of the gamut volume but leaving in-gamut colors more or less in place, exposes this profile funkiness. Perceptual rendering intent during conversion should scale inward all of the color values so that the most out-of-gamut colors get mapped to the volume surface - this will keep color relationships intact and, if the profile funkiness is something that is most prominent at the volume surface, will likely affect only a small number of pixels. Maybe this would explain why you are seeing the problem with relative colorimetric versus perceptual intent, because the rel. col. intent exposes the profile funkiness more effectively than perceptual.
Just a guess - I have admittedly crossed way over the boundaries of my color knowledge.
kirk
On Wed, Apr 15, 2015 at 12:41 AM, jeff.grant@... [QuadtoneRIP] <QuadtoneRIP@yahoogroups.com> wrote:
Using i1Profiler, it isn't an option, unless I've missed something. I wonder what they used to build your profile. I checked a few profiles and they are all Perceptual.