Improvements over the stuff you've been playing?
Certainly.
Enough to beat Magnus?
Maybe not.
Why do you mention Magnus? That's going to the extreme.
I'm talking about high regarded authors. When playing their recommendations and afterwards trying to analyze as thorough as I can it seems now and then they miss the best moves.
I don't think that authors claim to necessarily identify best. There main focus tends to be on what has been tried, with, from time to time, a suggestion for an improvement.
Authors even write false information, as Kochnoi did in ECO to trap unwary opponents. But basically, open theory is like clothes: players follow current trends.One must learn to think for himself.
I dunno, depends on what your "improvement" is.
Sometimes lines are chosen for practical reasons, and engines (or you on your own) will notice they "give up" equality or "give up" a small edge to the opponent.
Of course the author may be wrong if they're saying a line is drawing, but the line they offer is overly difficult, or saying a line offers good chances but it's borderline lost.
Most of books are written by 2500 humans. No doubt, there are tons of mistakes by 2500 humans in comparision with 3500 engines.
But even among engines books or without books still matters. ( cos without book, engines can analyse positions for a few seconds or minutes and still do mistakes ). There are books analysed by those 3500+ engines Leela and Stockfish for months and some books can be downloaded for free.
https://zipproth.de/Brainfish/download/
What do you think.