Error tolerance - that's exactly what the relevant people at CASA don't understand. There were no real tolerances specified last time I checked!
(Though I may be a little out of date.)
There must be some hidden away in the exam computers because I understand the questions now require numeric answers, but that's not how our mate at CASA thinks. He genuinely thinks that you follow the "correct" procedure (which only he knows) then you'll get the correct answer. Our mate genuinely thinks "tolerances" are how many decimal places.
If you can convince CASA to do something sensible like specify "you must be able to plan a B727 fuel burn to within +nnn/-nnn kg for flights of up to xxxx nm, under the following conditions xxxxx" then the industry will sing your praises. It would be easy enough, representative of actual FPL requirements, and conform to government policy!
So to answer your question - acceptable "tolerances" seemed to be:
1) record and use to the nearest 1 (1 nm, 1 kg, etc), and
2) be sure you use the "rounding" rules in the exam guide for temperature, wind, etc.
3)
do not interpolate except as specified in the exam guide. That will give you a more accurate answer but they want to see you follow the SOPs.
If you follow the rules for selecting wind level, temp deviation, etc, and apply correct alternate requirements, you should get the correct answer.
(PS. I didn't find much advantage to combining multiple legs into longer, averaged legs - unless the long leg is obviously same heading and weather. It saves a little time but opens up more chance for error. But that was just me.)
Good luck.