Analysis Engine forgets we are human

I have real grief with the chess.com analysis I feel personally that it doesn't count for the fact the people are playing are human beings. It produces lines of following that people rated under 1300 would Never in a thousand years would have followed or come up with. In this particular case  I would doubt that many would turn down a free piece when offered.

I am not referring to my infamous bishop sacrifice on f7/f2. But totally unrealistic lines.

So it came up that the best move here should have been Bc3 - I understand that. 1 move later

1 move later

I moved the Bishop were previously it had said it was the best move, now it is saying Bb4 is best -  No one would resist taking that bishop with the rook. 

22...Rhd8 23.b3 Kf7 24.Bxd6 Rdc8 25.Qxa7 Ra8 26.Qb7 Kg6

Now my point here is, this is totally unrealistic line of following - and we are getting scores based on this type of screw ups.

I do have couple of Chess computers and just for giggles I put this game in there everytime I played Bb4 they screamed at me disastrous!

I can't really say this isn't a bug because it is doing what it is designed to, which is why I'm posting instead of bug reporting. 

First Bb4 will run to Qd6+ and Qxb4 is not possible due to Bf8

After Bg7 Rxb4 runs to Qxd6+ and Qxb4

YOU ARE NOT SO ACCURATELY

Maybe just because it is computeranalysis? I guess you should refrain from seeing it as a mentor and more as a way to identify your mistakes.

In your example the situation for pinning the horse to the rook totally changed when your oponent played Bg7. 

And for the "no-one could resist...": take a look at some GM games. Possibly let analysis show you the lines if the Bishop was taken next move. It would cost black a rook due to a nice check/fork!!!

I don't think it's unreasonable, you're luring the rook where you can fork it with check. So if rook takes the bishop on b4, then queen takes pawn on d6 and can then capture it and you'll have a rook for a bishop is protected by a tactic.

It is weird though that it shows different results for the same move, I had something similar happen as well where one move showed an alternative but playing the alternative it showed that that alternative would be an inaccuracy.

Bc3 is best Bb4 is superior.. With some tactics black can fall for.

That fact explains why they are a 1300 player. Would you rather get guidance from a 1400 player or a 2400 player?

Studying the best move and seeing why it is superior often is only 2 or 3 moves deep. One goal in studying computer analysis is to improve one's move consideration. Having the view "I would never hang a bishop" will limit the moves considered and miss clearance and triple attach motives keeping one's rating stuck where it is. 

There may be glitches and hacks

I said here it is accurate

(Come to the point)

I guess thats the point of analysis, to show you moves you would never have thought of.

Why analysis needs support players rated 1300? It helps you identify the best moves to play better, or will you just want to stuck at 1300 level?