Wood and co-workers (Cavanaugh and Wood 2002; Robinson and Cavanagh 1998; Wood 2005; Wood and Murray 2003) have pioneered the application of some statistical methods to the taxonomy of various biological organisms, an approach that has become known as statistical baraminology (SB). For example, in Wood (2005), the report of an analysis of patterns of morphological characters in 30 turtle species, there appeared to be two distinct groups. From a statistical perspective, however, these methods are flawed in several ways, as was argued in an earlier paper Reeves (2019), where the question as to whether a conventional statistical approach—cluster analysis—could provide a more securely based alternative was left open. Part 2 of this research, reported herein, presents a formal reanalysis of the turtle data using some well-known clustering techniques. The results suggest that the original conclusion in Wood (2005)—that there are two distinct clusters—can be supported, as well as the detailed cluster composition. Nevertheless, clustering techniques, being more soundly based, should be preferred. In addition, they also enable the identification of how well individual taxa fit into their clusters by means of silhouette plots. Some further suggestions are made for the evaluation of cluster stability that have a sounder statistical basis than the type of bootstrapping commonly applied in SB. Finally, answers to some important questions concerning the original dataset are still needed, and these are used to highlight some important conceptual and practical issues within SB research.