Human-to-human transmission of the melioidosis bacterium, Burkholderia pseudomallei, is exceedingly rare, with only a handful of suspected cases documented to date. Here, we used whole-genome sequencing (WGS) to characterize one such unusual B. pseudomallei transmission event, which occurred between a breastfeeding mother with mastitis and her child. Two strains corresponding to multilocus sequence types (STs)-259 and-261 were identified in the mother’s sputum from both the primary culture sweep and in purified colonies, confirming an unusual polyclonal infection in this patient. In contrast, primary culture sweeps of the mother’s breast milk and the child’s cerebrospinal fluid and blood samples contained only ST-259, indicating monoclonal transmission to the child. Analysis of purified ST-259 isolates showed no genetic variation between mother and baby isolates, providing the strongest possible evidence of B. pseudomallei human-to-human transmission, probably via breastfeeding. Next, phylogenomic analysis of all isolates, including the mother’s mixed ST-259/ST-261 sputum sample, was performed to investigate the effects of mixtures on phylogenetic inference. Inclusion of this mixture caused a dramatic reduc-tion in the number of informative SNPs, resulting in branch collapse of ST-259 and ST-261 isolates, and several instances of incorrect topology in a global B. pseudomallei phylogeny, resulting in phylogenetic incongruence. Although phylogenomics can provide clues about the presence of mixtures within WGS datasets, our results demonstrate that this methodology can lead to phylogenetic misinterpretation if mixed genomes are not correctly identified and omitted. Using current bioinformatic tools, we demonstrate a robust method for bacterial mixture identification and strain parsing that avoids these pitfalls.