Precision astrometry from the second Gaia data release has allowed astronomers to predict 5787 microlensing events, with 528 of these having maximums within the extended Gaia mission (J2014.5-J2026.5). Future analysis of the Gaia time-series astrometry of these events will, in some cases, lead to precise gravitational mass measurements of the lens. We find that 61 per cent of events predicted during the extended Gaia mission with sources brighter than G=18 are likely to be spurious, with the background source in these cases commonly being either a duplicate detection or a binary companion of the lens. We present quality cuts to identify these spurious events and a revised list of microlensing event candidates. Our findings imply that half of the predictable astrometric microlensing events during the Gaia mission have yet to be identified.