To improve the accuracy of Ritchey-Common test for a flat mirror, a new method to use the relationship between the system pupil coordinate and the test mirror coordinate to interpolate and fit the test mirror surface was proposed. On the basis of the least square method, the system defocus error and surface error in two test angles were detached to gain a more actual flat surface error. The simulation analysis shows that the test error can be controlled under λ/100(λ=632.8 nm). A flat mirror with a diameter of 40 mm was tested. In the test process, the test flat mirror was rotated with different angels and the wavefront was reconstructed through the relationship between coordinate mapping and amplitude conversion. The obtained results show that the RMS of the test mirror surface is 0.018 6 λ after detaching the defocus error of the system. As compared with the test RMS of 0.021λ from an interferometer, the residual error is 0.002 4 λ . Experiment results indicate that this error detaching method is valid and accurate in the Ritchey-Common test for flat mirrors.