Color deficiencies and color correction

Visicheck (www.visicheck.com) has a very cool Photoshop plug-in to simulate
deuteranopia. For me personally, Protanope gamma 3.03 produces almost no
perceptible change (on my monitor, likely with the calibration skewed, since
I eyeballed it personally). Very highly saturated yellows become a little
less so, but the difference is slight and the match exact otherwise. I
believe the difference can be accounted for as a limitations of the monitor
or the color space. Color samplers placed before applying the filter show a
midling increase in red and green and no change in blue. The numeric
difference is significant; the unchanged appearance can't be explained away
any other way. I'm satisfied that that the filter represents the error
between what I see, and what "normal" sighted people see.

The question now is, how do I apply this new knowledge? Is it portable from
image to image, system to system, and color model to color model? Should I
continue to eyeball things until they express my intent, and then finish by
applying the filter to "correct" it? Would it be better to develop a monitor
profile to account for the vision difference? How would I go about doing
that?

Thankfully, I don't do this for a living...