The Spectral image function used for sampling is:

The Matrix used for converting from XYZ space to RGB is:

This matrix was obtained from the book "Priciples of Digital Image Synthesis"
by Andrew Glassner, and is used for standard NTSC monitors with phosphors
and white point given by the following table:
 

	x	y
r	0.670	0.330
g	0.210	0.710
b	0.140	0.080
w	0.313	0.329

 

Here is the the spectral image function sampled with different strategies to push values outside
the gamut into the gamut:
 

The strategy for this image was to simply clamp the RGB values individually.
The strategy for this image was to take the absoulute value of the RGB and clamp anything greater than 1. This obviously leads to colors having to much intensity.
This strategy is based on the above, except instead of taking absoulte values I scaled all negitive values by -1/3. This helped desaturate the image to a more accurate representation, although still to intense.
This strategy found the max absoulute intensity in the image and divided each RGB by it. Also it clamps negitive values to 0.0
This is the same strategy as above except it finds the max intensity, instead of the max absoulute intensity. It's probably exactly the same as above...
This strategy found the min and max intensities in the image and set them to 0 and 1 respectivley. It then linerley iterpolated the remaining values between them.