Numerically they're both right.
For a flux I'd look at cell values, so a strip of cells. It's one of (the many) reasons I teach that planning the post processing is part of the model set up. So, if you know you need data at a certain position build that into the mesh if it won't cause problems at the meshing stage.
You are navigating away from the AIS Discovery experience