why is the analysis so often wrong

Analysis depends on depth. The lower the depth, the less accurate and at low enough depths, it can be completely wrong, since it can take more moves (depth) to actually evaluate correctly.

For example a depth of 10 only looks 10 moves ahead while a depth of 18 looks 18 moves ahead. I think the max for a computer is 40? 

Bonkers I say ! The tool changes it’s mind from one move to the next- tells you a move was the best and the next ply it becomes a mistake... or vice versa !

= depth 

The downside is players who think the accuracy % is an indicator of engine use and report games as cheating.  The quick analysis tool I heard from Danny long ago is being revised- but she’s still off her rocker. 
One move is a blunder- after the next move it becomes the best ! Not much can be changed as it’s completely depended on depth of search- except perhaps in the terminology used.

What everyone have said is true but sometimes players are mistaken when they think that a certain move by the engine is wrong. There are instances where there is tactics involved where you , for instance, sacrifice a minor piece but after 2-3 moves you get a rook for it or something similar. So, when you think analysis is wrong, my advice is to follow through a couple of engine moves and see if there is a deeper point to the first move engine suggested. 

Junk the tool. Is a brainstorm gone bad. Made for the beginner player but should not be the default option.

I think I agree with you @BadBishopJones3. The reason I’m not totally sure about my agreement is that given my level of comprehension and understanding, I cannot always “see” why my move might be a blunder unless I immediately lose a piece. So, I’ve learned that there must be something I’m not picking up yet and try to look further. Often times, I have to give up and move on. 

This might have been a subjective thing but I have a feeling that depth 18 was a bit better a few months ago. Now it can happen that the engine says I don't know +1 in the opening  for a pretty normal move. I think that this was rarely the case in the past or at least not as much as I see it now. I check the move later with the tool I have, and it says something +0.3 or 0.4, so pretty normal for the opening. As for straight up blunders, engine rarely makes a mistake. If it says blunder on probably any depth, it is certainly not a good move and you should try to see why is it a blunder. First go with the engine move and see why is it better, if it is needed go a bit further. If you don't come up with anything then make your original move and see what happens with engine line after it. In most cases you will be able to pinpoint why the move is a blunder. 

For more subtle mistakes, that is not always the case and it goes down to your chess understanding.

@willtitley -

Yes, an excellent and vexing question, and thanks for asking. Obviously with limitations on computer power and resources, the review engine can’t be right all the time. 

My view is that the engine is right more often than not, and I’m not sure to what degree, but I’d estimate maybe 3/4 of the time, so when it says I made a mistake or blunder, I try to look past my obvious objections and often I can understand its reasoning, sometimes not.

One common best move it gives me is to sacrifice a knight for the sake of forcing the opponent to double up on pawns. Now I completely understand that doubling up pawns is undesirable, but it doesn’t seem worth it to sacrifice a knight. Strange!

11. .. Qxa4 13. Nxa4 Ne2+ 14. Kh1 Nxg3

There's not going to be a big evaluation difference; that's why it's only a miss.

edit: actually, a full review strength it's considered a mistake 

Well, c6 might not’ve been a great move, but all I see is that by taking the bishop I’m losing my queen to the knight. What am I not seeing here? 🤔

It ends up with even trades and black stays up a pawn.