One force setting doesn't fit all objects
It's tempting to think of crushing or dropping an object as a calibration problem: set the gripper's force high enough to hold securely, low enough to avoid damage, and tune until it works. The trouble is that the correct force isn't a fixed number, it varies by object, and often by the individual object in front of the robot at that moment. A rigid, well-controlled parallel-jaw gripper calibrated to hold a firm plastic part securely will apply that same force to a ripe piece of fruit, a paper cup, or a thin-walled container and crush it, because the gripper has no way of knowing, from force setting alone, that this particular object needed less. Turning the force down to protect fragile items, meanwhile, risks dropping anything heavier or more slippery than expected. There is no single setting that is simultaneously safe for a fragile item and secure for a rigid one.
This is compounded by irregular geometry. Even a rigid gripper applying appropriate force can fail if the object's shape means contact is concentrated at a single point or edge rather than spread across a surface, since pressure, not just total force, is what determines whether a surface yields or cracks. An object that is delicate specifically because it's thin-walled or brittle at certain points can survive a distributed grip and fail under a point contact even at lower total force. Precise force control alone, tuned in advance, cannot fully solve a problem that depends on the object's material properties and shape at the moment of contact.