The use of opportunistic communications, while increasing in popularity, is still limited, due in part to some uncertainty that still exists regarding its performance in real-world conditions. This paper tries to assess the real performance of an opportunistic routing implementation in a physical setting, by comparing it with its expected performance, determined by a simplified theoretical model. For that purpose, we have deployed an experimental testbed combining static and mobile sensor nodes, and running two different applications in two different platforms in tandem. This allows us to obtain real contact traces from an unmodified application, while at the same time logging the messages transferred between participating nodes. The data collected was later analysed, mainly in what refers to the intra-contact and global communication patterns, as well as the end-to-end delay distributions for each sending node's messages. The results obtained show that the system behaviour can be predicted with high accuracy by our simple model.