Geomorphological mapping is an important tool in geomorphology and related disciplines, yet it depends on the expertise and experience of the practitioner. The reliability of the technique and its products has not been subject to sufficient quantitative evaluation. In this study, we evaluated the magnitudes of differences in mapping glacial cirques between five maps in a mountainous area in the Southern Carpathians in Romania and attempted to identify the causes of the differences observed and possible solutions for obtaining more objective geomorphological mapping. We found notable differences between maps (in paired sample comparisons in all cases for the total values in each dataset) in the number of cirques, the total area and in headwall crest lengths. Statistically significant differences were found between datasets based on different semantic models of glacial cirques. Differences in mapping arise mainly from differences in conceptualising glacial cirques. When mapping relied on an explicit semantic model (a geomorphometric approach), differences were significantly smaller. Therefore, explicit semantic models of landforms based on land surface variables can result in more similar maps and further facilitate the transition from manual delineation to automated recognition of landforms from Digital Elevation Models (DEMs).