You are right; the results will be like this as long as a Cartesian mesh is used. This can cause significant discrepancies, especially for flows around streamlined bodies and curved surfaces. Basically, we cannot create these in Discovery Live right now (in either 2019R3 or 2020R1).
I'd like to make it clear, that in my opinion, the interpolation of the results which cause the flow field to appear unrealistic near the large gradients (where the spatial resolution is not sufficiently fine) happens only in post-processing. I think the results are not inherently corrupted in this sense; it's just a visualization (smoothing) effect, like switching between cell centroid and node values in Fluent for contour plot visualizations.