When book moves disagree with engine BEST moves

For example, in one of my Slav games:

https://www.chess.com/analysis/game/live/20813425413?tab=report

Engine likes 4..dxc4 best, in fact likes it better by about 0.6.

Yet the most popular book moves in this position (in descending order of popularity) are 4..Bf5, e6, followed by dxc4.

What accounts for the discrepancy?

The discrepancy is that there are very few games where White plays Bf4 because it is not accurate - as the engine makes painfully clear with dxc4.  Using Chess.com's master database, there are only 54 games with White playing 4.Bf4, and Black has a good win percentage after that! 

I think purely what happened with 4... Bf5 being the most "popular" move just has to do with accuracy from the players.  Additionally, most of the games were played before 2015.  You will notice that the games where Black (properly) played 4... dxc4 resulted in a good amount of success for Black.  

In short, my explanation is that 4.Bf4 is not a good move, is played seldom at a high level, and did not receive an accurate response the few times when it was played.

Makes sense, thanks for sharing.

It's because a position can be reached via different move orders. The numbers in the database mean that the position after Bf5 was reached in more games, but it doesn't necessarily mean that it had to be reached via this exact move order.

If you look at the games (https://www.chess.com/games/search?fen=rn1qkb1r/pp2pppp/2p2n2/3p1b2/2PP1B2/5N2/PP2PPPP/RN1QKB1R%20w%20KQkq), you will see that many games with the position after Bf5 started as London System and transposed to a Slav, something like d4 d5 Nf3 Nf6 Bf4 Bf5 c4 c6

In your position 4...dxc4 is definitely the most precise move.

Openings are very complex with many possibilities. Engines are often wrong due to the horizon effect: they cannot reach a meaningful evaluation at their calculation depth. Books are often wrong as well. The book summarises the practice of grandmasters, but as openings are so complex, in some cases many grandmasters have followed the same trodden path that leads to the abyss. When some human or engine refutes that line, then the books must change.