I've been making a few proofs on different papers. One thing I noticed is that some samples tends to be a bit more yellow or bit more magenta. This shows up in the soft-proofs as well so I think ICC profiles, although generic, are at least somewhat accurate. So my question is why do these color casts even remain in a color managed workflow? I get that paper white is going to be different but shouldn't the midtones be the same. Is it an issue of the paper white influencing how I perceive the colors in the picture? some other factor?
When using color management and printing photos, the chosen rendering intents are typically Relative colorimetric rendering (also with black point compensation commonly checked as well) or perceptual rendering. It's called relative rendering not for nothin'
Relative rendering maps the colors relative to the paper white LAB value, and that relative relationship flows well into midtones and even shadows. Thus, pure gray values in the source image file will map on the cool side of pure neutral gray when printed on a cool white paper, while mapping to warmer grays when choosing a warmer white paper.
Perceptual rendering is similar, but automatically includes the black point compensation and usually tends to lift the midtone luminance values a little as well to get slightly better midtone contrast compared to a straight relcol rendering. That said, not all profiling apps offer a proprietary perceptual mapping that differs from the relcol mapping, so there's that issue to contend with as well. You can confirm whether the perceptual tag is a proprietary one by checking both rendering intents with softproofing in Photoshop or other soft proof enabled app. If the perceptual tag is proprietary, you'll see a small but noticeable difference when comparing the RELCOL choice to the perceptual choice.
Lastly, you'd get more consistency between warm and cool white papers like you were expecting using Absolute colorimetric rendering, because it attempts to get as close as possible to the original colors, which sounds like what everyone would want, but it's not that simple, either!. It's only used by a very few printmakers for printing specific images because one has to manually bring both the whites and the blacks into the paper's printable gamut, especially the L* values, otherwise ABSCOL rendering intent will typically try to add some unwanted ink in those areas to force, for example, the cool white image areas more neutral, or the warm white image areas more neutral in color appearance. Generally, that's not what folks want to mess with because when its not done correctly it comes at the expense of the paper's maximum whiteness and black density response, thus, ABSCOL is rarely if ever recommended as a useful rendering intent when printing photos.
cheers,
Mark McCormick