why multiply the angle by the lightness?

[makemeunsee]

https://github.com/jradavenport/cubehelix/blob/master/cubehelix.py#L99
Here you use the "fract" array to build the angles array, but "fract" contains the interpolation from minLight to maxLight, and not a simple [0..1] interpolation.
I'm not sure it's a bug but it seems weird enough to raise the question.

[jradavenport]

This is a valid point!

This goes back to the original Fortran code that D. A. Green published with his definition of the cubehelix: http://adsabs.harvard.edu/abs/2011arXiv1108.5083G

[makemeunsee]

No, it's not in the Fortran code. In the Fortran code, the FRACT values always span from 0 to 1, by steps of 1/(NLEV-1), and does not allow to specify a minLight nor a maxLight like in the port. With specified values for minLight and maxLight, the angle value is not making sense anymore in term of hue (0 will not be blue, nor 1 red, etc.) and the behavior will not match that detailed in the documentation of the function.
See my (untested) pull request for a possible fix.