Should the engine play perfect moves after the first blunder in a puzzle?

Many times when I am doing puzzles I am surprised to see that after I react to the first move the engine sometimes isn't doing the BEST respond and may times even a very poor one..

is it supposed to be like this? shouldn't it do the perfect moves reacted to the situation?

The "best" move is a move which guarantees you the best score. There is no difference between checkmate, +64.9 and +4.3, they all win - they are all perfect. Same for lost and drawn positions.

When you are in a lost positions it is bad practice to keep on looking for the "best" move since there are none. It's better to set a trap inviting your opponent to make a mistake. The puzzle engine is set up to test your skill on the level of the puzzle rating. It knows it is lost anyway and it plays a defense which challenges you on one of the many different chess tactics.

Sometimes the engine does worse moves in a puzzle because 
1; it might not have been something you would have thought of, and therefore YOU have higher chance of getting it wrong
2; it doesn't matter the engine is already losing

To me I think the first one is more correct. The engines already know they are losing and the puzzles are preset to make it harder for HUMANS. Winning a piece yet? No not yet, I will attack your queen first. (idk how engines think)
Maybe engines are trying to get you to go wrong. 

Actually, I don't know if the solution moves are totally automated or approved by a human controller. Sometimes the engine plays 1 stupid defense, sometimes 3 and sometimes none. The only logic I recognized is that it sometimes makes you find a quick checkmate even when not necessary to win the game. For the rest it will stop interrogating you once there is more than 1 simply winning move (>+3.0 evaluation score)

i hate covidteen

ten