One of the challenges for bioinformatics is predicting protein complexes in protein-protein interaction (PPI) networks. A wide range of methods are used, each with a different approach and varying degree of accuracy. A common characteristic of current approaches is the generation of a large number of false positive predictions. This largely reflects the difficulty for the methods themselves to accurately distinguish true complexes from noisy or incomplete PPI network. Building on this observation, this paper presents a visualization inspired by enrichment maps. This visualization offers a perspective on protein complexes, highlighting clusters of true and false positives and revealing patterns in the predictions. We apply four representative prediction methods to three different PPI networks and examine how their predictions relate to established reference complexes. By highlighting systematic differences in prediction methods and focusing on clusters of similar results, including true and false positives, our analysis provides new insights into how these methods behave. These insights complement our previous work on structural asymmetries in PPI networks, contributing to a deeper understanding of how to evaluate and interpret the results of protein complex prediction methods.