I've done a bit of peer review in my area of expertise over the years. Peer review kind-of relies on the fact that what you're reading in the paper is honest. A reviewer can't be expected to recreate the results - the peer review process is intended to ensure that the paper's conclusions could reasonably be drawn from the results that are presented and the reported interpretation isn't simply fiction. If the reported results are incorrect, that's very difficult for the reviewer to ascertain. There are many reasons for reported results to be "incorrect". Typos in spreadsheets, confirmation bias, commercial interest, the "publish or perish" model that obermd has already mentioned . . . simple dishonesty . . . I'm sure you can think of others. Peer review is not perfect . . . but it's way better than the "no peer review" that some pay-for-publication portals allow.
Treading on the toes of giants . . .