Thank you for doing this! While it didn't give me the whole answer, the video together with what you wrote here gave enough info for me to seemingly figure it out.
What I understand so far:
1. Forget diagonal 4c prediction. No matter how you insert LR, opposite UL and UR colors before LR insertion mean it's the cycle case.
2. When inserting LR with one of them solved in the upper layer, a line on U means the same as in regular LR insertion from the bottom layer.
3. When inserting LR with one of them in the upper layer but mixed with the opposite color (M2 from the last case), a line on U is unreliable. You need to imagine flipping the center color, and going by the line that gives.
4. For the solved case, when inserting LR with one of them in the upper layer, there is no line, but there is a block in the M slice (may need to imagine flipping the center as in #3 in order to recognize it).
Phew. Is this right? And if so, is the much harder recog worth it vs just setting up LR on the bottom?