How does average centipawn loss work?

So played a recent game

1. d4 d5 2. c4 Nf6 { D06 Queen's Gambit Declined: Marshall Defense } 3. cxd5 Nxd5 4. e4 Nb6 5. Nf3 Bg4 6. Nc3 e6 7. Be2 Be7 8. O-O O-O 9. h3 Bh5 10. g4 Bg6 11. Ne5 N8d7 12. Nxg6 fxg6 13. f4 Bh4? { (0.74 2.13) Mistake. g5 was best. } (13... g5) 14. g5 c5 15. Bg4 Qe7?! { (2.29 3.10) Inaccuracy. cxd4 was best. } (15... cxd4 16. Bxe6+) 16. d5 exd5 17. Nxd5? { (2.88 1.13) Mistake. exd5 was best. } (17. exd5 h5 18. Be6+ Kh8 19. Kh1 Rae8 20. Qc2 Kh7 21. f5 Kh8 22. Qg2 Ne5) 17... Nxd5?? { (1.13 3.51) Blunder. Qxe4 was best. } (17... Qxe4 18. Bf3 Qd4+ 19. Qxd4 cxd4 20. Bd2 Nxd5 21. Bxd5+ Kh8 22. Rac1 Nb6 23. Bxb7) 18. exd5 Qd6?! { (2.98 3.83) Inaccuracy. h5 was best. } (18... h5 19. Be6+ Kh8 20. Qd3 Kh7 21. f5 Kh8 22. d6 Qd8 23. fxg6 Rxf1+ 24. Qxf1) 19. Be6+ Kh8 20. Qg4 Nb6 21. Rd1? { (3.71 1.89) Mistake. Qxh4 was best. } (21. Qxh4 Nxd5 22. Bxd5 Qxd5 23. Be3 b6 24. Qg3 Rae8 25. Qf3 Qc4 26. Rf2 Rd8) 21... Rfe8?! { (1.89 2.91) Inaccuracy. Rae8 was best. } (21... Rae8) 22. Be3 Nxd5?! { (2.77 3.73) Inaccuracy. Nc4 was best. } (22... Nc4 23. Bc1 Nb6 24. b3 Bg3 25. Bd2 Rxe6 26. Qxe6 Bxf4 27. Qxd6 Bxd6 28. Bc3) 23. Bxd5 Rxe3 24. Bxb7?! { (3.71 2.68) Inaccuracy. Qxh4 was best. } (24. Qxh4 c4 25. Bg2 Rd3 26. Rxd3 cxd3 27. Qf2 Rd8 28. Rd1 d2 29. b3 b5) 24... Rg3+? { (2.68 5.11) Mistake. Qb8 was best. } (24... Qb8 25. Bg2) 25. Kh2?! { (5.11 3.99) Inaccuracy. Qxg3 was best. } (25. Qxg3 Qc7 26. Qxh4 Qxb7 27. Rd2 Rf8 28. Kh2 Qc7 29. Rf2 Re8 30. Qg3 h5) 25... Rxg4?! { (3.99 5.30) Inaccuracy. Qxd1 was best. } (25... Qxd1 26. Qxd1 Re8 27. Bc6 Rb8 28. Qe2 c4 29. Bg2 Rd3 30. Kh1 Rbd8 31. Rc1) 26. Rxd6 Bg3+ 27. Kg2 Bxf4+ 28. hxg4 Bxd6 29. Bxa8 h6 30. Rd1 Bf4 31. Rd8+ Kh7 32. gxh6 gxh6 33. Rd7+ Kg8 34. Rxa7 { Black resigns. } 1-0
   
   https://lichess.org/g8uq5rkQ/white#0

Its fairly instructive in how effective putting pressure is, but also that complicating the position doesn't mean only your opponent will suffer- as can be seen in the dips from the game analysis

But, the review shows

So I just wanted to see how the avg centipawn loss comes out to be 27. Only taking the value of the 4 major dips for ease of calculation, has the advantage drops of 2.9->1.1 (-1.8) , 3.7->1.9 (-1.8), 3.7->2.7 (-1) and 5.1->4 (-1.1)

So just these 4 moves have a combined drop of -5.7 or a centipawn loss of 5700. This being a 34 move game, the average should be >5700/34 should it not? But the review only gives 27, way less then the 168 that should be contributed from just these 4 moves.

What am I missing?

I think centipawn loss calculates of how overall flow of Game was good( it caclculates from move to move difference from evaluation between best engine move and yours) , so it means how Strong Your Game wasю It's paramets, that shows how Strong was Your own Game
and Accuracy shows how much is Your Game similar to how would engine play

Hhrrmmm....

5.7 × 100 = 5700