Biodiversity databases increasingly function as the default substrate for species distribution mapping. Many processes can decouple archived occurrences from field reality, but one of the most consequential yet structurally omitted is documented local extinction—evidence that a population of a particular taxon previously recorded at a locality has since disappeared. We argue that local extinction is not simply absence but a directional, high-information transition (presence → loss) that, in database terms, requires data extirpation—something most biodiversity infrastructures cannot represent natively. The result is systematic temporal inflation: contemporary range products retain legacy presences, obscuring real-time decline and propagating error into conservation assessment, species distribution modelling, and invasion inference. To make biodiversity mapping time-aware without replacing occurrence backbones, we propose treating documented loss as a first-class record via a minimal, standardisable loss_event linked to historically occupied localities, with traceable provenance and (where available) a lightweight cause field. We anchor the proposal in an operational implementation using European crayfish extinctions, demonstrating that even fragmentary loss documentation produces measurable inflation in IUCN range metrics and revealing practical design constraints for spatial matching rules. This structural change—small in implementation but fundamental in effect—becomes essential as biodiversity informatics scales toward automated synthesis and AI-driven decision pipelines.