Luma curve. Manipulated photos to find manipulation. Correct.
You haven't shown that the artifacts you're attributing to tampering are anything more than an effect produced by your manipulation. It has been shown how the mathematics of your method produces them.
I included two free easy ones. I have several and use them to fact check the result of one.
You've provided two images you say are your own and whose provenance you can attest to. You say you have tampered with one and left the other alone. You purport that only one is revealed as tampered with via your method. However, the one you say was unaltered and should have produced a negative result is nonprobative because any "halo" that your method may have revealed is conflated with the high-luminosity effect produced in the background. This does not prove the validity of your method.
Your other attempts at experimental control simply use convenience photos you found online. If your method produces an effect, you declare the image to have been tampered with. If it does not, you declare it unaltered. This is circular reasoning and therefore not probative of the validity of the claims you make about your method.
@Andy_Ross has provided six photos, three of which are known to have been tampered with and the other three are known not to have been tampered with. Please apply your method to the photographs and show that the method produces a positive result on the tampered-with images and a negative result on the unaltered photos. Please show that your method can identify exactly in what way, if any, the image has been altered.
See the guy dancing. He says the left ine is AI and the right one is him. Both ai. Both controlled by computer. Rotoscooping likely. Notice the feet on the right. No shadow. No natural lighting. Not the lighting nature gives.
The provider of the images has said that one image is AI-generated while the other is a photographic image. But you seem to have determined that both are artificial. The photographic image is oversampled. We've already shown how your method has the effect of detecting edges in oversampled images. Unsurprisingly, it does so also here.
You claim there is no shadow associated with the dancer's foot and that this constitutes "unnatural" light that indicates a feature paste. This is naive. It's perfectly natural and expected for a cloudy day; all the lighting in the image is consistent with that condition. Further, one of your adjustments to the luminance reveals a faint shadow under the dancer's foot that is consistent with an ambient-occlusion type of shadow. Your expectations are not sophisticated enough to determine authenticity by eye.
*disclaimer. I am not interested in going back and forth about a dancer. If you disagree. Do so in one or two sentences. be direct. Concise. I will address it. Please. Do not do the 10 quote breakdown about this dancer. Do it to my event photos.
If you're unwilling to discuss your claims and evidence to the same level of rigor as your critics, you stand a very small chance of convincing them.