A previous submission, confplot (by Michelle Giuliano) plots a single solid error boundary at one level of confidence. I just built on that idea by allowing the shade of the boundary to vary in a graded fashion across an array of confidence levels taken from a t distribution. If the confidence levels are spaced closely enough, e.g. [0.01:0.01:0.5], then the confidence boundary looks very smooth. You can use whatever color map you want to get all kinds of nice-looking effects.
I can see how this would be useful. Part of the beauty of confplot is it is so ridiculously intuitive. If you could modify your code to handle some of the issues with confplot and make all this other stuff optional, I think that would make a lot of people happy.