The effect of measurement errors in the S-matrix of a reciprocal 2-port device is recognized in the (usually low) difference between S 12 and S 21 , as the device is nonreciprocal. This "false nonreciprocity" is analyzed in this paper, and it is verified that, for low loss device, the difference acts principally on the phases of S 12 and S 21. This anomaly can be removed if a numerical correction is applied to the experimental S-matrix. In doing so, it is proved that the residual measurement errors have comparable amplitudes on all scattering parameters. 2. Hence, at this stage, the question is to choose the best solution.