Visual simultaneous localization and mapping (VSLAM) is widely investigated for airborne applications, but fewer works focus on underwater VSLAM. Previous studies of state-of-the-art VSLAM in the underwater field demonstrate that while some stereo approaches are robust to underwater visual conditions, monocular ones still lack robustness to such case. The only monocular VSLAM system able to give partial but promising results in these studies are DSO and ORB-SLAM, but these methods are still limited by tracking inconsistencies or failures from which the SLAM system fails to recover. However, recent work extend the capabilities of these approaches in place recognition and tracking failure recovery. These new developments should therefore lead to better performance in underwater conditions. This paper presents an update of previous qualitative assessments by adding recent developments of monocular DSO and ORB-SLAM. The methods are evaluated in 8 underwater scenarios, considering three criteria: the percentage of the sequence for which a localization is estimated, loop closure detection success and map and trajectory consistency. The results show the interest of multi-map approaches, namely ORB-SLAM3, in improving significantly SLAM robustness to underwater challenging visual conditions.