capella-scan disposes of intelligent strategies to "sort" a large orchestra score. The challenge is to correctly identify staves in the System Template although all instruments may not be playing in each system. If this is not correctly done in capella-scan, then it will become impossible to create a voice extraction in capella. Via Edit - System Template... you decide which properties are considered for identification.
With a piece for flute and piano it will be sufficient to select "Bracket" and maybe "Clef". When flute and clarinet are playing together it is useful to add "Key" as identification criterion (as the clarinet transposes); "Clef" does not help in this case. The larger the orchestra the more criteria you will need to achieve a reliable allocation.
However, when you are scanning single voices please select no criterion at all - remove all ticks! This way you will achieve that recognition errors, for example at the beginning of a staff (clefs or accidentals) do not change the System Template. This way you can spare allocating staves during postprocessing.