Engine accurate?

One of the hot takes from the 2021 World Championship Match is that Carlsen's natural ability overcame engine prep from Nepo. Given the relatively low error rate from both players early in the match, I'm not convinced that this is true, although it is fair to say that Carlsen obviously did better when it wasn't just prep v prep.

Looking back at a number of older (pre engine) games is that their are often errors in analysis. The main reason for this seems to be that a lot of strong moves are overlooked at the top of the tree, either because it isn't part of an existing plan, or the assumption that the opponent is restricted to one or two choices. However, there are a number of examples where human analysis holds up against engine analysis, including the following game which I found in The Life and Games of Mikhail Tal.

Ravinsky,Grigory Ionovich - Ilivitzki,Georgi A [B93]
URS ch sf Riga (6), 1952

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. f4 e5 7. Nf3 Qc7 8. Bd3 Nbd7 9. O-O b5 10. a3 Be7 11. Kh1 O-O 12. fxe5 dxe5 13. Nh4 Bd8 14. Qe1 Nc5 15. Bg5 Be6 16. Nf5 Nh5 17. Qh4 Nf4 18. Nxg7 Kxg7 After the game a draw was found if Black played ( 18... Ncxd3 19. Qg3 Bxg5 20. Qxg5 Qd8 21. Nxe6+ ( 21. Qh6 Kh8 22. cxd3 Rg8 23. Rxf4 exf4 24. Nxe6 fxe6 25. Qxe6 ) 21... Qxg5 22. Nxg5 Nxb2 ) 19. Rxf4 exf4 20. Qh6+ Kg8 21. Bf6 Bxf6 22. e5 Nxd3 23. exf6 Nf2+ 24. Kg1 Nh3+ 25. Kf1 Bc4+ 26. Ne2 1-0