On using Summary Statistics from an External Calibration Sample to Correct for Covariate Measurement Error