In my opinion, PTE shoul behave in a similar way as Photoshop (PS).
In PTE we have nested structures using the Parent-Child-Construction. In PS we obtain nested structures of layers using (nested) Groups, and we have a kind of correspondence between the PTE-Parent-Objects and particular PS-Groups. It seems to me that PTE could use the PS blending mechanism. Regarding blending, groups of PS seem to be rendered independently if you avoid "Pass through". Furthermore, blending of parent objects in PTE and blending of the corresponding groups in PS should be the same. Perhaps, as a first step, you may study the situation where all groups and all parent objects have mode "Normal".
Applying this method to your example above, Img2 would be multiplied with Img1 (overlapping parts) inside the Parent-Child-Group, but not with the black background image. It would just cover the background, provided that the blending mode of Img1 is "Normal".