The Federated Learning (FL) framework has been applied in multiple domains, offering solutions that provide both accuracy and data privacy protection. Yet, specifically for human mobility prediction, prior solutions have been analyzed on mobility datasets that are spatially and temporally sparse, AND neglected the impact of the heterogeneity of users' mobility patterns. Heterogeneity on the (fine-grained) spatial and temporal mobility patterns directly impact prediction, hardening the FL performance analysis. As such, prior evaluations of FL on mobility prediction are limited and may overestimate the robustness of the proposed solutions. We here aim to fill this gap by analyzing the impact that different mobility patterns (e.g., repetitive and/or exploratory patterns) have on the performance of FL-based human mobility prediction models, in terms of both model effectiveness and efficiency.