In isolated oceanic islands, colonization patterns are often interpreted as resulting from dispersal rather than vicariant events. Such inferences may not be appropriate when island associations change over time and new islands do not form in a simple linear trend. Further complexity in the phylogeography of ocean islands arises when dealing with endangered taxa as extinctions, uncertainty on the number of evolutionary ‘units’, and human activities can obscure the progression of colonization events. Here, we address these issues through a reconstruction of the evolutionary history of giant Galápagos tortoises, integrating DNA data from extinct and extant species with information on recent human activities and newly available geological data. Our results show that only three of the five extinct or nearly extinct species should be considered independent evolutionary units. Dispersal from mainland South America started at approximately 3.2 Ma after the emergence of the two oldest islands of San Cristobal and Española. Dispersal from older to younger islands began approximately 1.74 Ma and was followed by multiple colonizations from different sources within the archipelago. Vicariant events, spurred by island formation, coalescence, and separation, contributed to lineage diversifications on Pinzón and Floreana dating from 1.26 and 0.85 Ma. This work provides an example of how to reconstruct the history of endangered taxa in spite of extinctions and human-mediated dispersal events and highlights the need to take into account both vicariance and dispersal when dealing with organisms from islands whose associations are not simply explained by a linear emergence model.