It seems that the problem could be with trying to align each of the entire complexes. Predictive models of structures with multiple protein chains can have a difficult time with finding the correct interactions between chains.
Have you tried the 3D alignment with each of the corresponding 3 chains? I took the A-chain of 7cr5 and did a 3D alignment with the B chain of 4rR0 using Sheba in Yasara and got a Calpha RMSD of 1.094 A over 111 aligned residues with 100.0% similarity. I have attached a PDB file of the aligned structures.
Best regards,
R.J. Richardson