Niche partitioning is the most studied factor structuring communities of competing species. In fragmented landscapes, however, a paradox can exist: different taxa may competitively dominate different types of habitat patches, resulting in a form of spatial niche partitioning, yet differences in long-term distributions among species can appear surprisingly small. This paradox is illustrated by an emblematic metacommunity - that of Daphnia spp. in rockpools on the Finnish Baltic coast, where three species compete with each other, have distinct ecological preferences, yet largely overlap in long-term distributions. Here we examine how metacommunity models that explicitly estimate species-specific demographic parameters can solve the apparent paradox. Our research confirms previous studies that local extinction rates are influenced by environmental variables in a strong and species-specific way and are considerably increased by interspecific competition. Yet, our simulations show that this situation exists alongside interspecific differences in realized niches that are, overall, small, and identified three main explanations for this compatibility. Our results illustrate how state-space modelling can clarify complex metacommunity dynamics and explain why local competition and niche differentiation do not always scale up to the landscape level.