As a result of the widespread adoption of RAD-Seq data in phylogeography, genealogical relationships previously evaluated using relatively few genetic markers can now be addressed with thousands of loci. One challenge, however, is that RAD-Seq generates complete genotypes for only a small subset of the loci or individuals. Simulations indicate that including loci with missing data can produce biased estimates of key population genetic parameters, although the influence of such biases in empirical studies is not well understood. Here we compare microsatellite data (8 loci) and RAD-Seq data (six datasets ranging from 239 to 25,198 loci) from red mangroves (Rhizophora mangle) in Florida to evaluate how different levels of data filtering influence phylogeographic inferences. For all datasets, we calculated population genetic statistics and evaluated population structure using principal component analysis (PCA). For RAD-Seq datasets, we additionally examined population structure under a coalescent framework using SVDQuartets. We found higher FST and FIS using microstatellite datasets, but that RAD-Seq-based estimates approached those based on microsatellites as more loci with more missing data were included. Analyses of all RAD-Seq datasets resolved the classic Gulf-Atlantic coastal phylogeographic break, although this was not significant in the microsatellite analyses. Applying multiple levels of filtering to RAD-Seq datasets can provide both a more complete picture of the potential biases in the data and stronger support for even subtle phylogeographic patterns.