There are four options of working with duplicates in Recovery Manager for Exchange (RME):
1. Allow duplicate items to be created
2. Replace all duplicates with restored items (restore takes longer)
3. Replace duplicates only if restored items are newer (restore takes longer)
4. Do not restore duplicates (restore takes longer)
The first option is the simplest one. RME does not care what exists in the production Inbox and just copies everything it finds in the backup. No burden besides copy operation.
2 to 4 options use another approach.
For each message found in the backup, RME retrieves its PR_SEARCH_KEY property and does a search (!) in a target folder. If a duplicate is found, RME applies its 'remove duplicates' logic (compares creation times, replaces if it's newer, etc - depends on the option selected).
The question then arises: "What affects the recovery speed?"
1. Which 'duplicate removal' option is selected (1 is faster, 2,3,4 are slower);
2. Number of messages in target folders (a big number of folders with fewer messages works faster than one huge Inbox);
3. Speed of MAPI connection between RME (Outlook) machine and target Exchange Server.
If, for example, a better performance is obtained while restoring to another mailbox, could be explained by fewer messages in the mailbox and/or by another folder structure.