In mobile ad hoc networks (MANETs), how to measure communication anonymity is a crucial issue. In our previous work , a theoretic approach based on evidence theory was proposed with detailed analysis. However, localization errors and scalability issues were not considered in the system assumption. In this paper, we further develop our work to incorporate localization errors in anonymity analysis. We propose the concept of super-nodes to model group based mobility. Time domain is sliced into intervals. In each interval, our proposed approach categorizes mobile nodes into clusters based on a novel metric that integrates geographical distances, historical distance records, and communication hops. We then provide the algorithm to generate super-nodes based on cluster formations from each interval. Evaluation results exhibit a satisfactory accuracy to recover group formation using super-nodes.