Analysis is objective, within the limits of of an engine's horizon and depth. In most positions, a line that ends up with even material, without a positional disadvantage, is going to show that in the analysis. If your line only works if your opponent doesn't play the best, then the engine is going to flag that as bad, because it is.
You should never hope your opponent won't see a refutation. Regarding the human-ness of explanations, that is a relatively complex thing to get right by looking at engine output from an algorithmic standpoint. The site has staff working on things to explain better, but it certainly isn't trivial.
I kinda understand how the analyses engine works, though I have some suggestions.
Currently its able to tell and identify certain good things, like getting your peaces active, connecting rooks on the 7th line and so on.
What it in my opinion fails to see, is that we humans create traps, and sometimes have a bit longer plan, exchanges are not always that bad. The analyser tells its bad to loose a peace even if it is taken baken back within a few moves. And thats often how we humans play... and sometimes those plans dont work I admit, but it often can work as.
Chess engines usually know the best move, chess engines rarely offer peaces for exchange.
So if an exchange of pieces happen (bishop knight and such) just note if there is equality in within 3 moves or so.. then you can say :
"Tthat was a risky plan, but your plan worked you tricked your opponent"
"This exchange failed, you lost pieces here"
"My computer voice says this is wrong play, however you fooled your opponent here"
"That was tricky, you made it but could have got more out of it if"
"That was tricky you survived, you missed a better"
Going backwards in the analyses one might tell there is a strong structure at the board:
"That was a good move, now you control the center"
"that was a hard fight on the center, (sadly not) in your favour"
As for end games some indications might help like 7 moves before the mate
"your position becomes ideal for mate"
"that was a sudden mate, you might have surpriced your oponent here"
"it looked dimm, but thanks to an error you won"
"This is a great positional move, it came handy with the final mate."