About game analysis

In this game on move 11 i played Bxf7+ with the idea that after Kxf7, Ng5+  allows me to capture the bishop on g4. My opponent did not go for it but when i analyzed the game i noticed that the engine thought that Ne4 was better and only after that Ng5+.

However if you analyze the line the engine recommends it will then say that Ng5+ is actually the right move. Why is that ? Shouldn't stockfish be able to always tell the best move in such straightforward positions?

If Stockfish recommended a different move, then it was better at the depth it looked at, though it may not have been much better.

They're essentially equal moves at depth 20. 

By depth 23, Ng5+ is preferred

I see, what i found confusing was that it shifted recommendation without me changing the depth.  I only looked at depth 20, and it said that Ng5+ was only "good" at first but maybe it needed time to calculate?

They're so close in evaluation at that depth so any particular eval run could put either move as top.