be be running as base-current-controlled current sink when it's on. There's no guarantee that they'll have the same current gain, so they'll pull down one side of the transformer harder than the other. Presumably the circuit sorts itself out, but it isn't exactly surprising that it gets "squirrelly" under light loads.
Sometimes it's useful to separate the regulator from the power train - just to sort of get an indea of what the power train wants to do over the intended control range.
These show the basic requirements and also the limits of the thing you're trying to control.
I always get suspicious when I see control node voltage or current ripple at multiples of the conversion frequency, at the last point of control for a power train that is itself incapable of significant slew rates, and may require 'balance' for at least a full conversion period.
RL