How to improve the sub-aperture stitching interferometry accuracy is critical for the sub-aperture stitching system. For one-dimensional flat sub-aperture stitching system, the influence of the stitching stage positioning error, the reference surface error and random noise on the stitching accuracy is simulated and analyzed of data based on both double sub-aperture stitching algorithm and error averaging stitching algorithm. The simulation results show that, the flat stitching system is insensitive to the high order errors of reference surface and the random noise. The stitching error of the high order errors of reference surface is a little larger than that of the random noise. The low order errors (the second order errors) of the reference surface will be accumulated and enlarged during the stitching process, which is the main error source for the flat stitching interferometry. The stitching accumulation error of the low order errors of the reference surface cannot be effectively controlled by the error averaging stitching algorithm. The double sub-aperture stitching algorithm and the error averaging stitching algorithm almost get the same stitching result. A flat mirror with aperture of 450 mm×60 mm is tested by 15 sub-apertures. The deviation between the stitching result and the test result of a large aperture interferometer is reduced from λ/3 ［peak valley value (PV), λ=632.8 nm］to λ/45 (PV) after removing the low order errors of the reference surface.