The first problem you have is the fish are not actually counted at the locks. Counts are made and then expanded for times not counted, lockages not seen, and so on.

Probably the only truly real number is commercial catch sold to a buyer. That gets counted one fish, two fish, red fish, blue fish.

Sport is expanded from CRC. Escapement is expanded from index counts. Dropout is an estimate, based on averages from studies elsewhere. Pinniped predation, likewise, is an expanded estimate.

It should surprise nobody that a collection of estimates does not add up to another estimate.