Why Does this Post Process Mask Example Work?

I have found this example on Youtube. A character is masked out by checking the “Render CustomDepth Pass” on the mesh and subtracting the “SceneTexture:SceneDepth” node from the “SceneTexture:CustomDepth” node, as shown in the attached blueprint graph. My issue is: Since the custom depth render target is completely black, with the exception of the mentioned mesh, wouldn’t the subtraction of the custom and regular depths return only zeroes and negative numbers, thus only rendering from the “SceneTexture:PostProcessingImput0” ? The way I thought it worked was that the black pixels (zero values) from the custom depth minus any value from the scene depth pixels (positive values) would return only negatives and zeroes, so the clamp node returns only zero values to the alpha on its connected node. The non-black pixels of the custom depth (The pixels of the mesh) are identical to the pixels of the regular depth scene (Also the pixels of the mesh), so the subtraction would return only 0s. But the mask works, I tried it. Why does this work?

331476-regular-depth.png