Model composition can be viewed in model-driven engineering as an operation where a set of activities should be performed to merge two input models into a single output model. The latter aggregates syntactical and semantic properties from the original models. However, given the growing heterogeneity of model composition strategies, it is particularly challenging for designers to objectively assess them given a particular problem at hand. The key problem is that there is a lack of canonical set of indicators to quantify harmful properties associated with the output models, such as composition conflicts and modularity anomalies. This paper presents an inquisitive study in order to capture an initial set of metrics for assessing and comparing model composition strategies in two case studies. We apply a number of metrics to quantify different conflict types and modularity properties arising at composite class models produced with override and merge-based strategies. We have observed that some of the quantitative indicators were effective to pinpoint when a model composition strategy is not properly chosen. In some cases, the output models exhibited non-obvious undesirable conflicts and anti-modularity factors.