HI YuKai,
yes, it does not make a difference if you run the OpenFOAM simulation from CAESES or directly from the command line.
The only possible explanation that I can think of is that the two simulations converged at different states. For example, in one simulation the flow is attached to the geometry and in the other it is seperated. But that has nothing to do with CAESES and only depends on your simulation setup. Ideally, you want to have the same result every time you run a simulation with the same geometry and configurations. One solution could be to increase the mesh size to have a higher resolution in regions where the flow changes a lot (e.g. seperation points).
Best regards,
Simon