Apparently, in Wordfast Pro there is no safe way to tell if a translator forgot to translate a fuzzy match or if she did it in purpose.
We are not keen to make the default behavior consider a fuzzy match translated after a leverage, because in several workflows, some people use TXML bilingual files for reference as they work. We believe that showing unedited fuzzy matches as legitimate search results could confuse the translator in some cases.
We recommend acquiring the habit of adding and removing a blank space in a unedited fuzzy segment as a method to validate it. This way Xbench will be able to distinguish among forgotten fuzzy matches (which will be still unedited with a score lower than 100) and manually modified/confirmed by the translator.
Also, the trick of making a fake modification of the unedited fuzzy segment to change it to a modified state will also allow you (and your client) to upload the segment to the memory for future reuse as a 100% match in the next iteration of the document.
If you leave it as unedited fuzzy match, you cannot use the Commit Segment to TM command and the segment is not uploaded to the memory, so it will be a fuzzy match every time you process future updates of the document.