The reason the JPEG artifacts are relevant is because there are extreme levels of artifacting even in light portions of the underlying werewolf image. This would indicate that the underlying image was heavily compressed before the newer portions of the image were added. Additionally, the gamut range of the artifacted underlying image do not match those of the suspected overpainted portions, which results in portions of the image with similar color reacting differently to gamut changes in the Shadow/Highlight tool. The 'overpainted' areas, including the pasted-in moon, show little to no artifacting at all or artifacts that are not consistent with the 'base' image of the werewolf. Further, there are numerous areas outside the red box I highlighted that show artifacting that is duplicated exactly which shows poor attempts at either covering up unwanted details, or attempting to extend the piece with interior sections.

JPEG artifacting normally takes place in all color ranges. The JPEG algorithm uses nearby color sampling (while comparing to diagonal contrast) to compress data, more than it uses black areas to hide sampling data. (By comparison, MPEG makes extensive use of black values to store compression data, much more than JPEG).

If it were a scan of a completed piece, artifacting would be consistent throughout the range. Even if it had been started, stopped, then restarted, artifacting would be consistent. The inconsistency in both color gamut, compression and style provide evidence (not proof) that multiple image sources were used, that underlying art was heavily manipulated, and that the process was done inexpertly. I also suspect that the image was intentionally darkened to hide evidence of the manipulation. Gamut adjustment in the opposite direction (not shown) displays a clearer view of details obscured by the overpainting, in essence showing that the underlying image had clearly painted eye detail.

Regarding the noise I mentioned, the noise is only at the edges of the moon, and is consistent with color-based selection tool use prior to copy/paste operations. The moon image is highly likely a photograph based on comparison to the exact dimensions of lunar features. Presumably a moon image would be against a perceivably black background. Color based selection outside the moon, and then inverted, would result in a selected moon. Invariably there is sensor noise from high ISO photographs of the moon and this would result in unwanted selection areas (pixel sized) after inversion. After a copy paste operation these would be almost imperceptible except after extensive gamut range inspection.

It is not ironclad proof. He has already been incontrovertibly shown to produce art forgeries, copies, etc. That comparison wasn't intended to prove he painted over it (which was presumed) but rather to provide an insight as to how such deceptions can be discovered. Even the most experienced 'Photoshopper' tends to leave trace evidence of which operations were performed. I still believe the most obvious evidence, however, is his utter lack of knowledge in regard to proper use of light and shadow in regard to anatomical form. Most of his fakes display such amateur mistakes on top of professional work.