I think the best method with any engine is not full game analysis.
My preference is to go through the game one move at a time with infinite analysis. You will see a lot more of what you missed. Also, if you wonder why a move is correct, play your sequence out and see how it's refuted.
I can't tell you how many times I've spotted a line that ends with a fork (or such) and then missed a crushing tactic from the other side.
One or two printed out lines can't beat going through every single move with as much or as little detail as you need.
Infinite analysis is also a good way to "check the cheats" after a loss.
Usually, the blatant ones stay in the top 3 moves of the infinite analysis for most, if not all, of the game.
Occasionally, they will even play a move not on the list, and in the end, it comes out with a higher evaluation! (That means they're probably using a better engine or looking deeper.)
Just remember, don't become obsessed with numbers. Numerical evaluations are only one part of the evaluation of a position.
Even computers are capable of making numerical mistakes. I played a hippo one time, where the game was completely locked. The comp had a space advantage, so it evaluated it as up a pawn or two, even though it was a dead draw. These things happen, not to mention horizon effects.
Remember, use the engine to aid study and understanding, but don't rely solely on it. Try to find the mistakes with a real board and pieces. Form a few ideas in your own head before you go running to see what the engine thinks. Use it as a learning tool, not a crutch.