Molluscan benthic assemblages provide unique opportunities for understanding both spatial and temporal patterns of biodiversity. Species richness in the shell remains found at a site (i.e. the death assemblage) is typically several times higher than in the counterpart living assemblage, reflecting a complex history of settlement, dissemination and decomposition post-mortem. We used high-density temporal and spatial sampling (>37’000 individuals representing 196 taxa) of a shallow (5-8 m) nearshore sandy habitat off the coast of south-east Sardinia (Italy, Mediterranean Sea) to study the factors responsible for differences in the relative diversity of living and death assemblages. We found that death assemblages at all sites were considerably more diverse than living communities (1.5-3.5x more dead than living taxa after sample-size standardization), with 78% of all taxa solely recovered as empty shells, resulting in low live-dead agreement. By carefully filtering the raw data and combining them with habitat information extracted from the literature, we disentangled the major causes of this discordance and quantified their individual effects. Increased dead diversities could not be attributed to undersampling of the living community, but instead resulted from three phenomena of decreasing importance: the post-mortem, out-of-habitat transport of non-indigenous taxa (57% of dead-only taxa were allochthonous), the time-averaged presence of rare indigenous taxa (40% of dead-only taxa), and the likely local extirpation of a small number of species (3% of dead-only taxa). Our approach demonstrates how ecological inferences based on death assemblages can be improved by restricting analyses to demonstrably indigenous taxa, and highlights how mollusc shell remains can be used to provide information over both ecological and evolutionary timescales.