In measurement practice, the residuals in least squares adjustment usually show various abnormal discrete distributions, including outliers, which is not conducive to the optimization of final measured values. In this paper, according to the physical mechanism of deviation, dispersion and outlier of repeated observations, it can be seen that abnormal distribution and outlier are normal measurement phenomena, and weakening the influence of outlier is an incorrect research direction. Then, by revealing the advantages of functional model processing, this paper puts forward the error correction idea of using the approximate function model to approach the actual function model step by step, and forms a new theoretical method to optimize the final measured values, which greatly improves the quality of measured values. This is a new measurement theory idea that is completely different from mainstream robust estimation research.