Generally any difference comes from different level of simulation, and difference in motion planning logic.
VC’s motion planning is “general purpose” so the real KUKA controller logic probably differs in many special cases. KUKA.Sim uses KUKA’s own robot controller simulator (RCS module), so the logic should be the same as on real robot controller, or at least very close to it.
The VC motion planning and simulation doesn’t take into account e.g. dynamics of the robot, so a robot arm with high payload will move just as fast at the extremes of its reach as same robot without any payload. I don’t know if and to what extent the KUKA’s RCS takes this into account, but it probably does, as payload can be configured in KUKA.Sim. Further details like settling time for servos might be difficult to simulate and calibrate.
If you only run robots within “comfortable” subset of they capabilities, and only use basic motions without e.g. splines, blending, singularities, controller modes optimized for speed vs. accuracy etc., then you might not see much difference between the generic motion planning in VC vs. KUKA’s own.
If you use VC 4.5 Premium with KUKA OLP add-on and the same RCS for motion control, then it should be practically identical to KUKA.Sim.