Units and dataset used by SunRGB

[nyorain]

Hey, first of thanks for this well-written library, and even releasing it into the public domain!
It has been really useful to me so far.

The SunRGB function gives unexpected/unplausible results for me. In an atmosphere with low turbidity, a sun at high elevation (e.g. cosTheta = 1) should have an illuminance of up to 100k lux (see for instance the measurements on page 37 of https://seblagarde.files.wordpress.com/2015/07/course_notes_moving_frostbite_to_pbr_v32.pdf). Using the reference implementation of the hosek wilkie sky model, I get these results (e.g. linear srgb (95k, 78k, 59k)) while using SunRGB returns rgb values that are usually around 2-4 times smaller (e.g. I'm getting something like (26k, 23k, 20k)). It looks like this function is supposed to return lux in linear srgb color space as well, where does the difference come from? I know that the reference implementation returns sun and sky radiance (and I'm converting that to illuminance), but the difference isn't that big since sky illuminance for this small solid angle isn't nearly that huge.
Am I just misinterpreting the values from this function? Could something be off with the dataset?