It's funny you should bring this up, I've been interested in being able to generate an accurate-looking NES palette, based on what my Panasonic CRT television's colors are. I use one of the palette testing homebrews on my PowerPak to do this.
Using a straight-up YIQ -> RGB matrix will give you values that are out of RGB range. The biggest problem are colors 22 and 23, since they're the most out-of-range.
On my TV, color 22 looks like a brilliant cerulean, and color 28 is a bright marigold. Color 08 is the darkest color in the NES palette (a brown that's nearly black). Colors 3C and 2C are cyan, but 1C and 0C turn bluer and bluer as it gets darker. Color 26 is strange, it's supposed to be a bright light-red color, but it's extremely easy to make it too pink, or too orange.
Of course, just using a palette test program isn't going to give you exact accuracy, you need to play games in order to see what non-adjacent colors look like when placed next to each other. Lately, I've been using Goonies II to test some of these.
The problem I get when trying to use a YIQ -> RGB matrix is that color x8 is too dark. Color 18 on my TV is brighter than you'd expect, but trying to generate it with the matrix makes it too dark. 28 comes out too dark too, it doesn't look like marigold, but a dirty gold color.
Color 22 warps to purple-blue when you clip it, and color xC doesn't get bluer as it gets darker.
I figured I was just missing something, like, I don't know how to emulate the CRT's gamma curves properly, and I don't know how to clip the out-of-gamut colors (like color 22) without warping them, or causing them to lose apparent brilliance.
Heh, the funny thing is, the palette looks almost perfect when you turn the contrast (white level) way down, but then it's too dark.
Edit: Another phenomenon is that the colors in the 0x range don't mix. For example, the first three colors are straight-up blue (but with slightly different brightnesses), then there's a purple color, followed by 3 pure reds (again, with different brightnesses), and then the superdark 08 brown, and then there's a few dark pure greens, and then a nice blue (not quite cyan) from 0C. It's as though the hues break down into the most dominant color channel as the luminance goes lower. Using a YIQ->RGB matrix doesn't reflect this.