Hello Kat,
I know quite a bit about the A1 (see my website), but until now, I've never seen the schematics for the B1. Thank you for posting them.
I do know that the B1 was, like the A1, a Tim de Paravicini design. He, as you may know, is a very innovative engineer who rarely follows convention. Hence, this amplifier might seem somewhat "quirky" at a first look. But my dealings with the A1 have taught me to look beyond this...
The symmetry is also present in the A1 design. You could remove each half of the amplifier (replacing it with a current source or even a low value resistor) and it would work! Understanding that this is, in effect, two separate amplifiers will help to work out how to adjust those preset resistors.
Looking at the top half, the LTP uses CFPs (complimentary follower pairs), but this doesn't change the basic operation. The output is buffered by EF T16, then T18 is the VAS. In a "normal" amplifier, you'd find a buffer between this and the output devices, but not here. Actually, T18 combines with T20/T22 to form a CFP, hence D1, which is a Baxendell diode.
The bottom half is the same, the only difference being T17, which is an emitter follower. Thus, the bottom half is Darlington rather than CPF. Presumably this was done to save cost...?
Anyway, key to the adjustments is the feedback resistors R50 and R51. Note that NFB is taken from before the 0.47 ohm resistors - not from the junction. Without the preset resistors, there would be 0V at the base of T10 (and T13) because the input is held at 0V by R33, and therefore, no standing current in the output stage (and crossover distortion). So an offset is introduced so that NFB can maintain a small voltage across the 0.47 ohm resistors. In theory, this NFB action should stop the unit entering thermal runaway, but I offer no warranty
As you've found, it's important that the offsets match each other so that the net DC offset is close to 0V. So I wonder if one should measure the voltages across R64 and R63 with respect to ground? This would make sense were it not for the extra transistors, and note that these are not included in the feedback! Yet their contribution to the standing current should have an effect on the THD, so I would guess that on the production line, these would have been lined up using a distortion analyser as well as a voltmeter.
Looking at your numbers, they don't quite add up. About 6mA might come from R58, but 15mA is "missing" somewhere. Did you have a load connected when you made these? Perhaps these are just the errors caused by the tolerance of the 0.47 resistors...
I don't know if they would have gone to the trouble of matching the output transistors for this low-cost model. I've been told that they didn't do any selection for the A1, but they did re-label the devices to make them appear more "mysterious"!
Sorry this has been a long post, and sorry for not giving you an authoritative answer. But hopefully it's helped explain the circuit - as I say, it shares much in common with the A1, and there were quite a few things I missed until T de P got in contact and explained them to me. It might be worth contacting him (lookup "EAR", his company), but I'd understand if he didn't want to respond after 25 years. Another bet might be Jez Arkless (email on my site), but he's running a business and understandably doesn't have much time to give free advice.
If you do get an answer, do post it here - I'd love to know!
Best regards,
Mark