Transport models routinely anchor points of interest (POIs) at the centroid of their building or parcel. On large multi-building sites, this can misrepresent pedestrian access, distorting first-/last-mile transit access, wheelchair routing, and curbside drop-off precision. We quantify how explicit entrance=* detail in OSM shifts accessibility-style metrics relative to centroid proxies. The study uses a global, population/built-volume-weighted random sample to compare centroids against the nearest main entrance. We release the full toolchain that persists measurements between POI / entrance / centroid anchors and the nearest road network and transit stop.