Your understanding of "control" and "variability" in this context are incorrect. The engineer may specify the tolerance, but the fact that the dimension varies from the nominal (ideal) value causes the system to experience variation in response. Imagine a steel I-beam that has nominal dimensions and tolerances on each of those dimensions. The i-beam that is on the larger side of those dimensions is going to be stronger than one at the lower (smaller) side of those tolerances. Of course there are many factors that go into this, including chemistry, temp, defects, etc. But everything else held constant, the dimensional differences is going to result in differences in performance. This is the variability that is being described.