Microsatellite loci are ideal for testing hypotheses relating to genetic segregation at fine spatio-temporal scales. They are also conserved among closely related species, making them potentially useful for clarifying interspecific relationships between recently diverged taxa. However, mutations at primer binding sites may lead to increased non-amplification, or disruptions that may lead to decreased polymorphism in non-target species. Furthermore, high mutation rates and constraints on allele size may also lead, with evolutionary time, to an increase in convergently evolved allele size classes, biasing measures of interspecific genetic differentiation. Here, we used next-generation sequencing to develop microsatellite markers from a shotgun genome sequence of the sub-Antarctic seabird, the thin-billed prion (Pachyptila belcheri), that we tested for cross-species amplification in other Pachyptila and related sub-Antarctic species. We found that heterozygosity decreased and the proportion of non-amplifying loci increased with phylogenetic distance from the target species. Surprisingly, we found that species trees estimated from interspecific FST provided better approximations of mtDNA relationships among the studied species than those estimated using DC, even though FST was more affected by null alleles. We observed a significantly non-linear second order polynomial relationship between microsatellite and mtDNA distances. We propose that the loss of linearity with increasing mtDNA distance stems from an increasing proportion of homoplastic allele size classes that are identical in state, but not identical by descent. Therefore, despite high cross-species amplification success and high polymorphism among the closely related Pachyptila species, we caution against the use of microsatellites in phylogenetic inference among distantly related taxa.