The identification of outlier loci in multilocus datasets is a critical step for reliably inferring population demographic histories and for detecting signatures of adaptive selection, since the presence of just a few of them can lead to a biased estimation of genetic parameters. Here, we show the effect of four outliers out of 30 microsatellites on phylogeography and relationship between eight native Italian goat breeds. The presence of the outliers influenced the relationships between individuals as estimated by factorial correspondence analysis, and between breeds as estimated by Reynolds’ distances. The outliers also affected the neighbour-joining (NJ) tree topology and the computation of the putative genetic barriers separating the populations. The complete dataset supported the existence of two significant barriers: the first isolates Orobica from the surrounding northern Alpine populations and the second separates Girgentana from the southern and Sarda breeds. After removing the outliers, a single and different barrier was significant. It separated the Sarda breed from the continental populations. Our results give new insights into the genetic structure of several native goat breeds across Italy and provide a clear example of the importance of testing for the occurrence of non-neutral loci in population genetics studies. Four outlier loci demonstrated the ability to blur the effect of geography, hiding the natural genetic barrier that separates the Sarda breed from the continental populations, while highlighting the peculiar demographic and selection history of Orobica and Girgentana.