Recently, a new logarithmic mixed pixel classification method has been developed through the establishment of appropriate networks. Based on the fact that natural targets do not consist of equally distributed components, the Network Based Method (NBM) alerts the user for non-sampled endmembers in the image scene. In this paper, detection of misallocated endmembers in the hyperspectral space is investigated through the Network Based Method. Detection relies on the fact that misallocation of an endmember in the hyperspectral space affects its signature because the endmember includes spectral components from other endmembers, mainly from the one which is approached mostly. Three experiments were implemented and their results were compared with the Sum to One Constraint Least Square (SCLS) method's results. Experiments showed efficiency of the method to detect two endmembers with common components.