John Simpson and I are conducting a GIPS(R) verification and I'm reminded of Ronald Reagan's advice to trust, but verify.
On multiple occasions we've encountered situations where the returns are incorrect, as produced by the software. Sometimes it's because the formula is simply wrong. An example of this surfaced a few years ago during another verification where I discovered that the composite return methodology the client's vendor was using was invalid. Another time a major vendor's calculation was in error: I told the vendor they had a "bug," but they said that no, the software was doing what they wanted. And they were correct! Unfortunately, what they wanted the software to do was wrong. Another vendor had a known "bug" that caused errors whenever certain corporate actions occurred; for some reason they allowed this problem to exist for several years before addressing it.
Sometimes a verifier will offer their clients formulas that turn out to be wrong. This happened recently with a formula to do carve-outs. The client had used this formula for several years, only to learn (from us) that it wasn't correct. We have also encountered rate of return formulas, though appearing to be quite right, are in reality, quite wrong.
It's unfortunate that this occurs more often than it should. Clients pay their software vendors and verifiers lots of money, and expect accurate and correct information in return.