Species delimitation is at the core of biological sciences. During the last decade, molecular-based approaches have advanced the field by providing additional sources of evidence to classical, morphology-based taxonomy. However, taxonomy has not yet fully embraced molecular species delimitation beyond threshold-based, single-gene approaches, and taxonomic knowledge is not commonly integrated into multilocus species delimitation models. Here we aim to bridge empirical data (taxonomic and genetic) with recently developed coalescent-based species delimitation approaches. We use the multispecies coalescent model as implemented in two Bayesian methods (dissect/stacey and bp&p) to infer species hypotheses. In both cases, we account for phylogenetic uncertainty (by not using any guide tree) and taxonomic uncertainty (by measuring the impact of using a priori taxonomic assignments to specimens). We focus on an entire Neotropical tribe of butterflies, the Haeterini (Nymphalidae: Satyrinae). We contrast divergent taxonomic opinion – splitting, lumping and misclassifying species – in the light of different phenotypic classifications proposed to date. Our results provide a solid background for the recognition of 22 species. The synergistic approach presented here overcomes limitations in both traditional taxonomy (e.g. by recognizing cryptic species) and molecular-based methods (e.g. by recognizing structured populations, and not raising them to species). Our framework provides a step forward towards standardization and increasing reproducibility of species delimitations.