Ecological networks are quantitative, graph-based descriptions of ecosystems, consisting of compartments (trophospecies and nutrient pools) that exchange fluxes of nutrients or energy. Previous research pointed out how the model's design is a crucial task that can heavily influence analyses results, and how merging compartments for the purpose of comparing two or more different ecosystems can significantly alter the indices on which the comparison is based. All these works have been focused on the aggregation of trophospecies, whereas networks may comprise several nutrient compartments that may be lumped as well, either for lack of information or for comparison constraints.We show how the aggregation of these non-living compartments can have a greater influence on network analysis results than trophospecies clustering. This problem should on the one hand encourage modelers to make an effort to test the possible effects of aggregations, and on the other show how the role of non-living compartments could be very important in determining network dynamics.