Ontology matching is a key solution to deal with the semantic heterogeneity problem in semantic web and data integration task. Generally, due to the high heterogeneity of ontologies, the combination of many methods are necessary to discover mappings. Then, an ontology matching tool can be seen as a set of several matching modules; each of them implementing a specific method (e.g., terminological based, structural based, semantic based, etc.). Moreover, many aspects should be taken into consideration. For example, what is the interaction between the all modules working together. In this paper, we study and evaluate with OAEI datasets two main issues such as: (i) global-based vs. local-based matching approach ; (ii) impact of input mappings on the quality of matching modules. In these experiments, we demonstrate that in element and structure based matchers, our proposed global based methods outperform the local ones and are more stable.