cypher: Although I agree they are useless for illumination purposes, I've had good success with all the reasons I stated using (faked) HDR starmaps in Blender (which doesn't have the 'detail' slider that exists in TG). I might be wrong about how useful they are in TG.
Kadri: the problem with the pics is that they are not based on exposure but on magnitude threshold, which is a non-linear illumination function that amplifies the dimmest stars to make them more visible in low-dynamic range images. (or dims the brightest stars, not sure which

In addition, stars in reality are virtually point sources, which is how they would be represented in an HDR file. (might be an argument for some multi-pixel stars). But because the NASA pics were meant to be just looked at, they applied a point-spread function ('PSF') that blossoms the stars out based on how bright they are, to make them appear 'brighter' to the eye.
There is a way to inverse the point-spread function to collect those spreads and shove them back into a pixel with appropriately high brightness values (beyond what a jpg can handle) but that is very difficult guesswork if the original PSF is not known.
njeneb: thanks for the link, for general backdrops this program looks interesting.