Accurately dating when people first colonized new areas is vital for understanding the pace of past cultural and environmental changes, including questions of mobility, human impacts and human responses to climate change. Establishing effective chronologies of these events requires the synthesis of multiple radiocarbon (C-14) dates. Various chronometric hygiene protocols have been used to refine C-14 dating of island colonization, but they can discard up to 95% of available C-14 dates leaving very small datasets for further analysis. Despite their foundation in sound theory, without independent tests we cannot know if these protocols are apt, too strict or too lax. In Iceland, an ice core-dated tephrochronology of the archaeology of first settlement enables us to evaluate the accuracy of C-14 chronologies. This approach demonstrated that the inclusion of a wider range of C-14 samples in Bayesian models improves the precision, but does not affect the model outcome. Therefore, based on our assessments, we advocate a new protocol that works with a much wider range of samples and where outlying C-14 dates are systematically disqualified using Bayesian Outlier Models. We show that this approach can produce robust termini ante quos for colonization events and may be usefully applied elsewhere.