I noticed some time ago that the gain of the current-sense outputs are not the same for forward and reverse in the VNH5019 driver, so that the two directions need to be individually calibrated. Perhaps this is a related issue. VNH5019 current sense -- unequal gains fwd/rev
