A yield map represents the basic information for precision agriculture and a site-specific crop management. Its accuracy is extremely important. Yield data processing and error analysis are key approaches to improving the reliability of the yield map and minimizing the influence of error on the yield data. Based on the first two years' practice on a grain combine with a yield monitoring system, this paper contributes to the yield data error analysis and improvement of the yield map. Three kinds of main errors of yield data were described: cutter width setting error, filling time error, and delayed time error. The paper discussed the error occurence, recognition and processing. Errors were compared under different harvesting conditions of corn and winter wheat. The percentage of cutter width setting error data in the yield data set is less than 6%, the percentage of filling time error data in the yield data set is more than 10%, the percentage of delayed time error data in the yield data set is less than 1%.