It would appear as if you have an anomaly based on the geometry and restraint. In the real world the shaft would disengage on one edge as it bends but apparently the bearing contact is not letting the shaft move.
You might try modeling the lower half of a bearing and using non-penetration contact between the shaft and "bearing". It won't be true but might end up closer to actual stress.
Usually stress will be near the fixture area.
Here you have fixed near the bearing area, so stress will occur near this. (as happened in your case)
If you feel stress is less, check for option - per item or total & units.