Airborne laser bathymetry (ALB) is an efficient and accurate tool for mapping submerged environments, particularly shallow water bodies that are difficult to access with surface vessels. Modern ALB systems can achieve accuracies comparable to SONAR. However, multiple factors, including geo-referencing, water surface modelling, and range measurements, influence the resulting point cloud, making analytical error propagation challenging. Empirical evaluation against reference data is therefore essential, but difficult: ALB accuracy is typically in the low centimetre range, requiring reference data of equal or higher accuracy. Robotic total stations enable acquisition of underwater reference data for shallow water depths, e.g., up to 4.5m, with expected accuracies between 3mm to 10mm, depending on water depth, which approaches the inherent accuracy of ALB and limits the evaluation significance. In this study, we assess a UAS-based ALB data set from a mountain lake in Austria using reference planes and points acquired by robotic total stations. We separate the accuracy analysis into trueness and precision to isolate the effects of geo-referencing and water surface modelling from the intrinsic uncertainty of the LiDAR sensor. The results show that geo-referencing introduces the largest systematic bias, while the precision of the ALB data remains approximately 1 cm to 2 cm, even for submerged measurements. These findings demonstrate the high accuracy of state-of-the-art ALB systems and provide a framework for rigorous accuracy assessment in shallow aquatic environments.