In the following position, McDonald states that 14. d4! is the move that gives a large advantage to White.
r1bqr1k1/2p2pp1/pbn4p/1p1np3/P7/1BPP1N2/1P3PPP/R1BQRNK1 w - - 0 14
However, 14. axb5 wins not one but two pawns and without any unnecessary complications, such as opening the center up and allowing Black some freedom.
After using Stockfish to verify this, I wonder why (or how) GMs write mistakes in opening books, when they have engines to double-check. Why does this happen?
In such a case, I would consider an editing blunder. Can you give the moves that lead to that position ? I would not be surprised, for instance, if the comment was written with respect to the same line where ...Bb7 is played instead of ...Bb6.
r2qr1k1/1bp2pp1/p1n4p/1pbnp3/P7/1BPP1N2/1P3PPP/R1BQRNK1 w - - 0 14
In this position 14.axb5 leads to nothing but exchanges, but 14.d4 indeed looks very strong: 14...exd4 loses to 15.Rxe8 Qxe8 16.Bxd5, and if the Bc5 retreats then 15.dxe5 grabs a pawn with a discovered attack on the Nd5. White wins (at least) a good central pawn.
Writing a book is a long process: you study a variation, finally the variation won't enter the book because you found an improvement earlier, you analyse a position, then another one that's very similar but not identical, you check one with the software but maybe you forgot the check the second, then you have to send a draft of a chapter to your editor, meanwhile you are writing the next chapter, and the editor wants you to reduce the material by one fourth, so you come back to your text, cut this and rephrase that, and at the end you have a few days (at most!) to proof-read the final draft before it goes to print... Even for the most organized writers, errors will happen at some point.
Even without a computer, there is no doubt McDonald would find 14.axb5!+- in less than twenty seconds if given the diagram in the question. That's why I think this error is due to the production process rather than the chess process. It could be anything, from a comment meant for another section being copied-pasted to the wrong spot, to the typo Bb6/Bb7, or the wrong draft section sent to print.
It is sure pretty annoying for you reader when you meet such a mistake – at least, it keeps your 'scientific skeptical awareness' in alert when reading. But the author will be the most pissed off by such an accident: not only the message he wanted to pass about a variation is spoiled, moreover his name is associated to a 'ridiculous' assessment.
Because often priority is to sell as much books as possible and write them in as short time as possible. What's the value for the author to spend plenty of time to get rid of couple of meaningless errors? It has virtually no impact on the overall value of the product. People don't use so robotic approach. Well, 95% of lines may be checked with the engine and 5% doesn't have to be because author could be tired or already watching movie during writing. As long as people do something, they make errors in it.
