Reachability queries are fundamental for analyzing connectivity in large graphs, supporting applications in domains such as finance, social networks, and program analysis. Despite the variety of proposed indexing techniques, previous evaluations are often limited to narrow graph datasets and fail to provide comprehensive comparisons or in-depth analysis. In this work, we present an extensive empirical study of representative reachability indexes across diverse graphs and query workloads. We utilize a cost-performance metric that jointly considers index size and query time, providing a unified measure of practical efficiency. Our large-scale benchmarking covers 1,400 graphs and over 38,000 query configurations, revealing how structural and query characteristics affect index performance. Furthermore, we construct a decision tree model to uncover feature-driven method preferences and provide practical guidance for index selection under varying workloads and system constraints.