I would be strongly inclined to think that the two phase panel is tapped from the A and C phases rather than from the A and B phases; just a bit of screwy labeling, including hidden stuff like the way the breaker stabs/connection points are 'interlaced' to alternate phases going down the panel, would make it easy to mistake which phase the sub panel is tapped from.
However there are physically reasonable ways that you could actually see current flowing _between_ the various loads connected together, such that the supply from the mains to phase B is lower than the current circulating on the bus and between loads on the main panel and loads on the subpanel. Your comment about not having a 'true rms' meter opens the possibility of harmonic currents flowing from one load to another.
Even without harmonics, you can see this sort of effect. The classic example is a parallel inductor-capacitor circuit. Remember that current leads in a capacitor and lags in an inductor. If you connect an inductor and a capacitor together in parallel, then the leading current will balance out the lagging current, and the external current flowing through this circuit will be quite small. But the current flowing _between_ the inductor and the capacitor (circulating current) will be quite large.
See:
http://www.tpub.com/content/neets/14181/css/14181_33.htm -Jon
