I am using the bvals and bvecs files as supplied with the DWI data and have made no modifications to it.
Most likely those responsible for generating / providing the data fell for this gotcha just as we did. Indeed if you were to load these same data using an older version of MRtrix3, i.e. before we fixed this bug, the two errors would have cancelled each other out, and the FODs would be oriented appropriately.
However, I’m not sure if just flipping the left and right based on a visual inspection is the right way to go about it. Is there a way to ensure that the bvecs/bvals are consistent with the DW image and automatically do a flip if necessary?
dwigradcheck is intended as a data-driven check for such issues. It can never be “ensured” that the gradient table is correct, but it will make its best guess as to what any error may be.