You would think so, but no. The correction is only to ensure the resultant measured qubit is correct.

Grover's algorithm apparently relies on very small phase angles iterated very many times, which is beyond the capabilities of current forms of quantum error correction.