91% Accuracy? Are You Mocking Me?

This isn't the programmers fault though , you just have to increase the engine depth if you want a more accurate report... If you go to an analysis board, you'll see a little cog wheel in the top right corner next to the game board. There you can increase the engine depth, it will give a more accurate calculation but it will also take much longer. The programmers only implemented stockfish so I can tell you that it certainly isn't their fault. You could also download stockfish for yourself and analyse the game manually. Link for stockfish download: https://stockfishchess.org or https://github.com/official-stockfish/Stockfish for source code. Hope I could help.

Thanks.  To clarify, I'm not saying the "excellent" moves are losing moves given that distinction in lieu of winning or drawing moves.  These "losing excellent" moves are usually the second best move in a position in which even the "best" move loses, in other words a dead lost position.

Ohh ok, gotcha!

Speaking for myself, I hadn't noticed that option, so thanks for pointing it out!

You're welcome

So? That only further demonstrates your inability to properly use it. If there's only book moves, stop looking at the accuracy. I am sure you can figure that out... It lets you know how many moves there are that it didn't like, that's it. It doesn't tell you how many times you hung your queen and mate in one and it doesn't determine the quality of the game for you. If you use it intelligently then it's an amazingly valuable tool.

FOOLS

Sorry - but it is of no help to the vast majority of users that the tool targets - new players , beginning players looking for a quick and simple tool to check their games upon completion of their game. They should not be expected to know - much less proceed with all the extra steps mentioned to arrive at a reasonable evaluation. 
Players click on the quick analysis for two reasons -to view their Accuracy score, that of the opponents and to make a quick analysis of the game. 
Both are flawed and bugged and often to not repeat when ran a 2nd time. The totals of best, excellent, good moves often do not match, sometimes substantially, when the analysis and the Report totals are compared. In analysis- the evaluation flips from best move to blunder when the next 1/2 move is played due to depth of search. New players can not be expected to understand all the nuances. The tool is directed towards and made for use for the learning players. It only results in confusion. This is seen by all the misinterpretation of the Accuracy score- which experienced players know is a complete joke (as exhibited by the OP) and results in endless, unfounded accusations of cheating, wasting countless hours of resources for the site and members. The programmers created a monstrosity- we’d all be better served by the tool being junked and replaced with one that is useful and beneficial. There is no such thing as a quick and reliable analysis tool making an accurate assessment of an entire game. It is a brainstorm that could never work.

I have games with 98% Accuracy that include blunders and mistakes with other games having no blunders and a single mistake and Accuracy scores below 60% - both games in the 20/30 move range. I can semi- figure some of of the logic, but in the most part it is unexplainable. I’m very familiar with the process involved. New players can not begin to understand the data that’s presented to them - data that is primarily given for their use.

A man was walking along the edge of a steep cliff, taking great care not to fall over the edge.  For 99 out of 100 steps, he was successful.

I have observed changes made to the tools analysis over time that involve book moves. Gambit openings were always immediately evaluated as mistakes. Now- programmers have rightly changed the openings to reflect as being book moves. The issues that exists are not necessarily inherent in Stockfish and a shallow search - but are often embedded in a programmed response which translates and interprets moves into terms.

The cliff thing is important to note.  Sometimes the computer can find exotic ways to kick still-inevitable loss of material over its time horizon.  A telltale sign of this is when the evaluation keeps getting worse even though the defending side is repeatedly playing the "best" engine moves.

Especially when it comes to positional moves, you have to trust the classical or hypermodern principles you know (the latter are better, but few players actually know them), the engine simply won't be of much help.

Understanding an engines analysis requires a bit of practice and training , that’s for sure. The whole idea behind the tool provided at the end of every game - is a quick analysis that concludes with an # evaluation. Staff has made it clear - the tool is for the new players to check and see where obvious mistakes or blunders were made. It is not to be used as cheat detection.

Unfortunately that becomes primarily exactly what it is used for. Players learn the analysis is unreliable unless the obvious Queen was hung. They use it simply to look at the opponents Accuracy score after losing a game ... thinking it reflects engine use or not.

Good analogy. People just really need to learn how to interpret the evaluation and it is quite interesting. It tells you how many moves were of a certain quality and how many were bad or blunders. 

It does make sense. It might be using a different depth. 

Ha ha !

Ever cross reference the totals given in the Report with the totals in the analysis? (Add them up)

The do not match! Often by a substantial margin.

They do add up? Click on them and it shows you all of them.

Ha ha  ! 
No at the same depth. It’s bugged.

Show me.

Show yourself. Staff is aware of the tools problems and have been working on “fixes” for some time now.