Abbyy FineReader is far and away the best we've ever used. Nothing comes close in the space.
Tesseract requires piles and piles of training, which you may or may not have.
Does a clean photo of a document yield 100% accuracy typically?
Agreed. I use it a lot and it is really, really good as you say.
I have played around with Google's Cloud Vision API and it is the best OCR I have ever experienced - a big step up from my old poorly trained tesseract setup.