Peak search and analysis of the programs often associate peaks found in gamma-ray spectra of the radionuclides that are not present in the sample. Such radionuclides have to be excluded from the reported results. The dilemmas connected with reporting of the results near the detection limit are usually avoided by introducing a reporting limit exceeding considerably the detection limit. This approach has the drawback that small concentrations of emitters may not be reported, even though enough evidence for their presence in the sample is available. In this paper the statistical analysis of causes, which lead to manual elimination of radionuclides from the results of automatic analysis is presented for a set of measurements in our laboratory, covering a period of more than two years. Possibilities to improve the software to automatically detect and eliminate such results are presented as well. From the conclusions additional instructions to the operators on how to evaluate such peaks can be compiled.