The analysis is running incredibly quickly. It can't get very deep: it's probably like depth = 10 or something like that.
Analysis "best" move

I have 8 cores @ 4GH, I think that can handle an engine. I think it can handle an evaluation.
It was indeed set to 10seconds, I've set it to 5minutes, but I can't seem to reevaluate it. I'll try again tomorrow
But still, if it's a premium feature, I think it should do better than play okay move on the very first move of the "best" line.

The depth is 20, it doesn't run on your computer but on our servers instead so the strength is pretty consistent.

Thanks for clearing it up. I remember seeing depth = 20 on the self analysis board.
But then a okay response to the best move instead of the best response does throw me off.
I'm not sure how the engine work, but couldn't the played moves and best moves be analyzed with the default depth plus line length. So if the line is 10 moves long, to have it analyzed with the depth of 30, thus guaranteeing that all moves played in that line would be best moves if evaluating with the depth of 20?

The horsie ends up taking the rook in the "best" line, which wouldn't be possible if black had responded with the best move instead of a move that's almost an inaccuracy. On a scale from brilliant, best, excellent, good, ok, inaccuracy, mistake, blunder: ok is closer to the bad side of moves and shouldn't be the first move in the best line
This is really bugging me. Maybe it's my lack of understanding the position.
My knight was on d4 and got attacked by the pawn with c5. I played Nf5 which is a mistake. It says Nxb5 would have been best. Looking at that best line, the next move for black is c4.
How can this be the best move, if the next move in the evaluation is an okay-ish move instead of the best. The knight ends up taking the rook, but that wouldn't be possible with the best move.
What bothers me is how this can be the best move if very next opponent move is assumed to be not the best. That's like saying going for scholar's mate is the best move, provided the opponent doesn't defend against it.
It might of course very well be the best move, but how are we supposed to learn from the analysis if the best lines include blunders from the opponent or am I completely mistaken here?
That's why I suggested for the first best lines to be calculated as line length + depth.
With the default depth = 20 and line length = 10, the second move would only have a depth of 19 and the last the depth of 10, so evaluating them with the full depth of 20 it might get ranked worse. With a depth of 30, the last move would still be consistent with the depth of 20.
Or what else might be causing blunders or inaccuracies in the suggested lines?