20220412-2a3c30e8f6-win_x86 shows a scoring anomaly then saving loading and restoring previous ligand designs.
See the shared with scientists solution 'current scoring bug' for puzzle 2136.
It has a saved score of 21227, but you get 18646 when loading it.
restoring best gives me 21183.
When I restore credit best for that puzzle (only one track used) I get yet another score which is lower than the saved one.

Apart from the scoring bug there are other issues with the user interface in ligand puzzles that I will not address in this post. I urge others to stay on topic when replying to this post as well.