The colonisation of the Pacific is an important chapter in human dispersal for which chronological control is primarily provided by radiocarbon (14C) dates. In this context, the ability to reliably date shellfish is important because alternative dating materials, such as charcoal and bone, are typically highly degraded. However, the interpretation of shell 14C results is not always black and white because 14C is not evenly distributed throughout the marine environment, with estuarine taxa more likely to incorporate terrestrial sources of carbon. Regions where water has percolated through limestone bedrock provide an additional problem since ancient carbon is introduced into the estuarine waters. This "hardwater" has been put forward to explain old 3500 cal. BP results from culturally significant shells recovered from the site of Unai Bapot (Bapot-1) on the island of Saipan (Petchey et al., 2017). While arguments for (Carson and Hung, 2017) and against (Rieth and Athens, 2017) early settlement dates remain polarised, little attention has been given to the idea of change in the marine 14C reservoir over time, or to possible species-specific offsets in shell 14C. In this paper, we further develop a tri-isotope approach using 14C, ?13C, ?18O to identify carbon source. To investigate which shellfish are more prone to erroneous ages we have selected shell taxa that cover a range of nearshore environments commonly found in Pacific archaeological sites; including Anadara antiquata, Gafrarium pectinatum (both estuarine) and Tridacna (marine/reef). To test the possibility of change over time we extend the dating of the site beyond the earliest occupation layers to deposits considered to post-date the end of the mid-Holocene drawdown in sea-level.