This is the feature you can find in Trados Studio and memoQ.
"Weighted" should be paraphrased as "fragment-based."
My native language is character-based. So, the source segment is like this:
If "GHSYS" (exactly equal sequence) is in the TM, when the fuzzy match threshold is 60% or less, that TM unit is missed (because of character-based, non-sequential counting).
My request is:
When there are TM units containing the same fragment of more than, say, 10 characters, they are picked up, despite the "nominal" fuzzy threshold.