Thanks for the links, these should help understand what is happening, although the paper you reference is written obscurely (!!??!!)

I tried to overlay Raul's composite world image (mondo_ridotto_0p25.gif) with the night sky image (earthlights2_dmsp_big.jpg)

It was immediately clear that these images did not fit nicely. There are shifts between continents that are obvious and bad. I suspect that mondo_ridotto is less well registered, but we have to get into complex stuff like projections and spheroids before we can clear it up entirely.

I need to compare both to a properly projected world image...

In principle, I can get a product in a couple of days that shows the overlay of the two images, but there is certainly a waveband effect between IR and true VIS...