The idea that a robotic hand grasping an apple is a simple specialization of ball grasping seems to underestimate the complexity.
There is about an 80% probability that passing from a ball, with its predictable geometry, to an apple, with its natural irregularities, introduces significant new challenges.
For robots on production lines, for example, the variation in fruit firmness would require much more sensitive force sensors and real-time adaptation, which is a significant technical barrier.
I would put the chances at 75% that the hand's ability not to crush an apple results from a much more advanced design and calibration than for a simple ball.