Maybe I should have included the whole paragraph:

"And even when "truth" can be clearly defined, it is a concept to which natural selection is indifferent. To be sure, if an accurate portrayal of reality, to oneself or to others can help spread one's genes, then accuracy of perception or communication may evolve. And often this will be the case (when, say, you remember where food is stored, and share the data with offspring or siblings). But when accurate reporting and genetic interest do thus intersect, that's just a happy coincidence. Truth and honesty are never favored by natural selection in and of themselves. Natural selection neither "prefers" honesty nor "prefers" dishonesty. It just doesn't care."

He's talking about the "maps" that humans/animals may carry in their brains. These maps don't need to match the territory to be adaptive (I think your criticism of the quote hinges on how you would define "significantly"). But there's quite a bit of space where a "bad map" does not prevent adaptive behavior.

For example, some non-venomous snakes "copied" the color patterns of venomous snakes. It's still adaptive for animals to avoid all snakes with this coloring (just to be safe) without needing to know the truth about which snake is dangerous and which isn't. And natural selection is "rewarding" the non-venomous snake for lying about how dangerous it is.