The evolution of high-throughput targeted sequencing using universal markers has allowed for the generation of substantial sequence data pools from non-model organisms at cost-effective rates. These methods target areas of interest such as highly conserved regions and functional genes under the assumption of containing enough genetic variability to differentiate between species or even individuals. However, despite the relatively low cost when compared to whole-genome sequencing methods and subsequent feasibility for multi-sample population research, there is room for improvement. Recent population genomic studies using Angiosperms353, a universal probe set for 353 protein-coding genes found in flowering plants, found little gene recovery bias based on taxon, high recovery in segregating sites representing both between- and within-species distinction, and hitchhiking recovery of flanking non-coding regions. It is therefore possible to genotype multiple SNPs occurring in both coding and their flanking non-coding regions together as a multi-allelic “microhaplotype” for use in relationship inference of individuals and populations. In order to conduct relationship inference, we require large sample sizes and new methods for high-throughput targeted sequencing analysis. We describe our use of low-volume automated liquid handlers to prepare target-capture libraries from herbarium specimens at ⅒ volumes and compare recovery results to ½ volume library sequences. We also propose a “microhaplotype” bioinformatics pipeline through examples of results from two congeneric species pairs with a special focus on core-edge range and widespread-endemic population demographics.