This paper is concerned with the issue of reputation and the computation of group opinion. We argue that entities may receive both objective and subjective opinions, and distinguishing between the two is crucial for achieving more precise measures. Additionally, we argue that the group opinion about an entity α is not only influenced by the opinions that α receives (whether objective or subjective), but by the reputation of other entities that α is related to. As such, we propose a method that permits the propagation and aggregation of opinions in structural graphs, allowing the inference of more precise reputation measures through the description of both objective and subjective group opinion.