Recovery of evolutionary history and delimiting species boundaries in widely distributed, poorly known groups requires extensive geographic sampling, but sampling regimes are difficult to design a priori because evolutionary diversity is often "hidden" by inadequate taxonomy. Large data sets are needed, and these provide unique challenges for analysis when they span intra- and interspecific levels of divergence. However, protocols have been designed to combine methods of analysis for DNA sequences that exhibit both very shallow and relatively deeper divergences. In this study, we combined several tree-based phylogeny reconstruction methods with nested-clade analysis to extract maximum historical signal at various levels in the poorly known Liolaemus elongatus-kriegi lizard complex in temperate South America. We implemented a recently descrirbed tree-based protocol for DNA sequences to test for species boundaries, and we propose modifications to accommodate large data sets and gene regions with heterogeneous substitution rates. Combining haplotype trees with nested-clade analyses allowed testing of species boundaries on the basis of a priori defined criteria. The results obtained suggest that the number of putative species in the L. elongatus-kriegi complex could be doubled. We discuss these findings in the context of the advantages and limitations of a combined approach for retrieval of maximum historical information in large data sets and with reference to the yet formidable unresolved issues of sampling strategies.