OCR does still suck (but many people don’t notice the mistakes), especially the open source stuff. It’s a hard problem to solve. Not even Google and Microsoft can get it right if you try to use their OCR at a slight angle for more than a single sentence.