When you are manually browsing a long list of duplicates to find which one to delete, you MUST mark them first.
Just SELECTING them is not enough, if Similarity is still running and searching for new duplicates. When it finds a new pair that will be added HIGHER than the first item you had already selected, the whole selection shifts. The result is (if you do not notice the additional dupes) that you might unintentionally delete the wrong files.
Please make selections 'fixed', not influenced by new entries.
