Its strange I agree the way the netwok is all pass the hf and Uhf via the 52uf capacitor.
I doubt the reason was voltage of the capacitors, as in bi amp mode the 52 is bypassed the if that were the case it would be worst.
However, it could be an attempt to provide overall uniform group delay to the whole Mid, hf and uhf drivers.
They might have also found an anomally with the overall phase response and felt it was better to keep consistency.
The mid transition 2nd orderer filter of the 4343 is mistuned to minimise mutual coupling of the 2121/2231 and again this may have created a whacky effect on the 2121 which if not also feed to the 2121 and 2405 might have shown up as a side effect of the 2121 and 2405 were not part of that filter.
That probably is it , I recall I had a hell of a time with the crossover when I built my 4343. I did not have access to the schematic at that point (c Sept 1979) and Gary Margulius wrote me a nice letter back when I asked for information. As I recall bass/mid transition really sucked as did the rest of it until I diced the bass/mid filter in favour of bi amping. ( the Nakamichi blcok box crossover was about 6 db down at 320hz)
Capacitors back then were like wet sponges and I recall I actually blow up the bi polar 50volt capacitors in my prototype network. They popped like streamers. I started off with text book values after measure the impediance and then spent 6 months tweaking the chokes in the HF passband until until it voiced better than anything else on the planet.
Dang it I am on holidays, why am I sitting on this, no its a JBL holiday!
Hope that helps unlock the mystery.
regards
Ian