That's now fixed and the algorithm doesn't go back to the beginning of the file every time a chunk is read. Silly me.