Calculating the transfer values separately at each stage isn't technically incorrect but a different method that was used in the past before computers made counting STV elections a doddle. Technically you've used the standard Gregory method while that calculator uses Inclusive Gregory (which doesn't allow a ballot paper to increase in value as the count goes on; which the former does) - the latter is better, but almost impossible to do by hand.
Yes, that sounds very familiar to me now that you mention it. Thanks for the clarification! It's been too long since I did this stuff on the regular and I was never very good at it. When I looked at the newest output I got from PL (which appears to be identical to the newly-published ones), I was huffing and puffing because TimTurner wasn't getting elected until the final round and potentially in danger despite hitting quota in the first round. He didn't have any surplus to redistribute and no one's surpluses went to him so I thought it looked very odd compared to what I remembered.
That's the issue with that particular software - what it should have done is just elected TimTurner and not transferred any of his votes: what it did was wait until he exceeded the quota before it would try to transfer votes - that's a problem since it means that if if he had gained a vote in an early stage: rather than that vote skipping him and going on to its next preference it'd be added to TimTurner; and then all of his votes would be redistributed which isn't correct. Doesn't change anything this time, but that sort of thing could impact future elections. Unfortunately, there's no better (free) options that I can see