Where the numbers come from
How duplicate matching works — the pipeline, and what it does not claim.
Duplicate a file in Finder on an APFS volume and macOS does not copy the data. It creates a clone: a second file that points at the same blocks, and starts diverging only when one of them is written to.
Both files are real. Both report their full size. Together they occupy the space of one.
That is a problem for every duplicate finder, because the obvious arithmetic — five copies of 4 GiB, delete four, free 16 GiB — is wrong. Delete four clones and you free nothing at all.
With Detect APFS clones on — the default on macOS — DuoBolt asks the filesystem for each file’s clone identifier while it walks. Files that share storage share that identifier.
It then counts physical copies rather than files, and computes recoverable space from those:
Removing all but one copy leaves one physical copy behind, so what you get back is (physical copies − 1) × size.
A group made entirely of clones reads 0 B recoverable, even though each
file has a real size.
The selection footer shows two numbers whenever they disagree: how much you have selected, and how much is reclaimable. Both are true; the second is what the disk will actually give back.
Selection maths follows the same rule — a clone cluster frees bytes only when every file in it is selected. Leave one behind and the blocks stay allocated.
Time Machine local snapshots, cp -c, Finder duplicates, several backup tools
and most macOS installers all produce clones. On a working Mac a meaningful
share of “duplicates” are clones — which is why a competing tool can promise you
80 GB and deliver eight.
Where the numbers come from
How duplicate matching works — the pipeline, and what it does not claim.
Turning it on or off
Setting up a scan — under Performance.