A robot is more than its CAD model. It includes tolerances, wiring compromises, software assumptions, test damage, and the environment it has to handle. The model is useful, but it only describes the version of the system that has not met the real world yet.
The interesting work starts once a prototype moves. A first test makes the abstract constraints concrete: a cable pulls where it should not, a mechanism binds under load, or a sensor becomes unreliable in the exact condition it was meant to measure.
That is not a failure of the process. It is the process. Every honest test narrows the gap between a design that works in theory and a machine that can be trusted outside the lab.
The useful question is rarely whether a prototype worked perfectly. It is what changed once it was tested, and whether that change points to a better next build.