Differences in measured lambda between banks can be due to many physical things (it's not in the tune):
for example, engine is not symmetrical left-to-right or front-to-rear, each cylinder burns a little differently, each header runner is a different lenngth;
the 0.030 difference you're seeing in Lambda is small (e.g. compare 1.030 with 1.000), it's 3%;
when LTFT's are enabled, the PCM trims each bank to stoich (lambda 1.000) so you won't see as big a difference;
if in OL (LTFT disabled) you're seeing lambda 0.970-0.990, you would make B3605 or B3647 leaner by 0.010-0.030 (which is a very small amount, you might now overshoot).