Are you running the two locomotives uncoupled to see this happening? If so, that is not a good way to measure list sharing. When uncoupled, if the locomotives run at the same speed they are speed matched. Load sharing is different. When load sharing, the locomotives will run at different speeds in order to spread the load felt by the locos equally across the consist. So when the B unit is following, it's likely sensing that the lead locomotive is not carrying enough of the load so it slows it's speed down.
On the flip side, when the B unit is in the lead the trailing A unit is sensing that the lead unit is pulling too much load so the A unit is running faster (and appearing to run at the same speed as B).
This is all likely happening because the B unit is drawing more amps when running when compared to the A unit. Could be anything in the drivetrain, motor, etc. I have several locomotives that exhibit this behavior because of various amp draw differences. It all evens out when in a consist.
You can run the FMLC setting again on both if you want, and that may help ease the differences between the two. If you really want to experiment you can adjust the FMLC manually up and down a bit to see how that affects running in a consist. I don't typically recommend doing that, though I have done it to a few of my models to fine tune things. Nearly all locomotives will work perfectly just by seeing FMLC automatically.
Kevin