These sizes are used to code the information deriving from molecular fingerprints into a fixed size bit vector. As far as I know in the current version there is no way to know to which fragment correspond each bit. Moreover each bit can also codify more than one fragment due to the fingerprints collision events.