Rank amateur with the nanovna with no right to chime in here. But here's a guess. SWR is a scaler quantity and is a minimum when the MAGNITUDE of the impedance is closest to 50 ohms. But on the Smith chart, might the algorithm for finding the minimum be to find the point where the plot lies on the real axis AND is closest to 50 ohms? I can see this happen and I have heard of cases where the "closest" point on the Smith chart doesn't correspond to the lowest SWR. Hard to say not knowing the algorithm for finding the "closest" point on the Smith chart.
