Thanks, I see the right axis now. I now realize that this can’t be done retrospectively. The algebraic sum of each main is converted to absolute value, which is not the same as converting each individual measurement to absolute value. So this doesn’t work. Sorry. it would have to be done by removing the allow negative values and running for awhile.
This is how I understand it to be setup:
So in the earlier period when L2+L3 were correct, L2 was assigned to phase A and L3 was assigned to phase B, which resulted in both referencing L1 for voltage, as the method describes. You only need two mains CTs, so any of these should work when added together (allow negative values off):
main1 phase A
main3 phase C
main2 phase A
main3 phase B
main1 phase B
main2 phase C
The individual circuits are basically single phase, so they should be:
L1-L2 phase A
L2-L3 phase C
L3-L1 phase B
If you could double check that this is how you have it and it works, that would be great. I wanted to lay it out for the next person that needs to do a three-wire system. Please get back and confirm the accuracy of the results.
If you choose the first mains option above, main1, phase A and main3, phase C, with the extra input from removing main2 you could connect CTs to L1-A and L3-C on that branch meter and the sum should match that meter.