I think you are over complicating things and I doubt if JLH went to this much trouble in analysing his circuit.
Your assumptions about the voltage across R4 is correct - it is clamped by the 2Vbes of the transistors. I am not sure if you can say that most of Q4's current flows into Q3 and likewise, Q3's current into Q1, but some of it will at least.
If you want to go with a detailed analysis of the circuit, it might be easier if you break the feedback loop at node 11, connect node 11 to 0V and analyse it from there. With feedback every node changes with every other node and you can end up driving yourself mad