(Too low?) engine depth analysing your games

Position is +9.61 after blacks move (first picture) but when white does "best move" the evaluation is +2.72 (2nd picture).

Evaluation jumping up and down like this happens almost every game I analyse and I think it makes it very difficult to use the tool. Even though I am a beginner (currently 1080 rating) and use the analysis mainly to identify my largest mistakes rather than to find the very best move in every situation, the evaluation is sometimes so much off that I think the tool looses much of its value. 

Is this behaviour because engine uses of a too low depth? I understand you get higher depth with a Diamond membership (I only have Gold) but I think you should expect better than this when you pay money to get the analyse feature! Whatever the reason is, I think you should consider increasing the calculation depth (or whatever the problem is) for all memberships so that we get at least descent performance in the analysing tool. At least this was not what I expected when I decided to sign up for a paid membership! 

I mean there are other totally free tools with much better performance so it can't really be a cpu problem can it?

I would really like to see someone from chess.com to comment on this topic.

I had the idea that I could use the analyse tool to learn from my mistakes and see what better moves I could have played. But how can I do that when the engine changes it's opinion all the time?

Another example below (it's not like I have to search for these examples, they occure almost every game):

I play Knight c3, Engine says its a mistake and recommends Bishop d2.

I play Bishop d2 instead and engine tells me best move is Knight c3! 🙄

With this level of performance I think you should not even market the lower level subscription as "unlimited analyses"!

Or does it matter what platform you use? I use the chess.com app on an android tablet, are the calculations done local  on my tablet or in the "cloud"?

The engine is the engine. You'll get the same type of results at a given depth with a standalone engine on your local machine as well. Some positions need a lot more time and depth to really resolve what is going on. And sometimes, that isn't enough either.

I had a game, on a local analysis, where it marked a move as a blunder and suggested a better move/line. Go into that line and it determined it too was a blunder and suggested a completely different move/line. Tried that one, and it too was a blunder. It eventually made it back to the move I made, which was still a blunder in that case.

Engines have tree pruning that occurs, depth limitations, and some things are past the horizon. In some cases, you'll want to go to the deepest you can. For the cloud based analysis, you'll be limited by premium level how deep you can go, with a depth of 26 as a standard option and depth 30 max analysis that has to be manually chosen if you want it. That analysis, uses server time and CPU cycles, and is limited.

You can set self-analysis and let it run as long as you want, though that uses local client resources and you can set it to unlimited and sit on a position for as long as you want to see if going deeper makes any difference to the evaluations in complex positions.

I just noticed you said the app. The Game Report is cloud based as far as I'm aware and the local analysis uses client resources, but there aren't any configuration options for either in that case. I don't know what the default cloud depths are for the apps.