post your ICC profiles and we can try to see what the issue is.
What paper type are you choosing? This has a huge impact on the amount of ink laid down. You could try "Special 10" as it should be the highest saturation of ink using matte black (Special 5 is the same for PK).
Canon's heads are 600dpi, so you're looking at something on the order of:
Draft: 600x600
Normal: 1200x600
High: 2400x1200
For a "good quality" paper stock.
But there are a few more factors at play that muddy the waters a bit-- some lower quality papers will drop down to 300x300 on draft mode and take everything down a notch.
I generally use Normal quality on everything I can unless there is banding in the feed direction (which on a Canon is almost always down to the feed adjustment and not a missing nozzle). These numbers are also kind of hard to trust as being concrete proof of anything since the X direction value is usually computed by increasing the amount of passes-- basically you're running the printhead over the same spot twice, so 600dpi becomes 1200 dpi, but given the resolution of the encoder strip in this direction, there really shouldn't be 1200 or 2400dpi of actual addressable space, or at least the combination of the speed of the head and the firing precision almost definitely doesn't equate to a real world number, since you're almost certainly not hitting the exact same spot you hit in the previous pass (nor do you necessarily want to, since the ink mixing is meant to happen optically, not by function of say, a cyan and a yellow dot combining on paper to make a single green dot. In some ways it's like talking about digi cam megapixel values-- when you talk about a "24 megapixel sensor" you're not taking into account the bayer filter array that halves the resolution of the color information caputured, except in this instance you're also dealing with 12 inks, so it's even more complicated. I don't know the actual mathematical truth about how this all works, and I think it probably varies depending on your screening algorithm, and even the color you're trying to produce. For example, one square pixel in a digital file could potentially require 12 different dots to create at the correct color value, but does that now make the output resolution 12 dots per n value?
What it boils down to, same as with Epson, is do some testing, and find the setting that works best for you. In my opinion, don't just jack up the settings to the highest because it "must be better". But yeah, also I agree that it's all confusing, but my point is kind of that having number values wouldn't necessarily simplify it.
Oh and going back to your issue with the color profiling, this kind of HAS to be a profiling/rendering intent issue unless the printer just can't hit that part of the gamut (although it sounds like it kind of can given your workaround). This is assuming that you're using the same profiling software and gamut mapping settings therein as you did when you made your Epson profiles.
