CST EM Studio： Symmetrical geometry but asymmetrical results

I seem to notice that a circularly or cylindrically symmetric problem geometry in CST EM Studio sometimes results in field plots that are noticeably asymmetrical.
IOW, a simple magnetostatic problem involving, say, two disks of permanent magnet material (and BTW, how accurate can real-world permanent magnets be simulated if the only specification for the magnets is a "magnetization vector"?) spaced a diameter or two apart should give rise to axially symmetric field plots, both vector and scalar, I would think.
Yet not only do the field plots seem somewhat asymmetrical, but subsequent force calculations don't seem correct either. IOW, the problem is symmetric about the Z axis, but the Z directed forces on each magnet are not equal, differing from each other by 20% or so, while the forces in the x and y directions, while much smaller than the Z-directed forces, differ by two orders of magnitude for one disk magnet, and 60 or 70% for the other magnet.
Moreover, I've noticed the same kind of errors in Particle Studio. An acquaintence wanted to build the Van de Graff type electrostatic particle accelerator as per plans detailed in "Scientific American" magazine, and I attempted to model the accelerating column consisting of six equally spaced drift tubes between the HV terminal and ground, but for certain geometries and voltages where the beam would strike the inside of a drift tube, the beam did not symmetrically strike the inside of the tube. Rather, it was noticeably skewed to one side, even though the geometry was axisymmetric.
I'm wondering, has anyone else noticed this, or is it just me doing something wrong?

hello,
I have had similar problems in cst mws. the results became better (=more symmetrical) when a denser mesh was used.
best regards

For the case of two identical permanent magnet disks facing each other along the Z axis, the default mesh properties for some reason caused the mesh in one disk to have 50% greater mesh density than the other identical disk.
And I'm finding that making changes to the mesh settings makes a big difference in the field plots and force calculations, and I don't see why it should, especially for a simple, symmetrical problem geometry with adaptive mesh refinement turned on.
For example when I vary the mesh properties and run the problem, I get widely varying magnetic force results; almost a 2 to 1 variation in fz.
edit:
I found out that when the adaptive meshing "refinement strategy" is set to "energy based", the mesh is nonuniform, the field plots are asymmetrical, and the numbers are way off.
When it's set to "expert system based" the mesh is uniform, the field plots are reasonably symmetrical, and the forces are equal, but it never seems to converge; the more the mesh gets refined, the smaller the B field and the forces get.

To the best of my knowledge, sometimes the same problem shows up for Q3D of ansoft.