Someone explained to me that transcribers work under instructions to write what they see and NOT to -interpret- what they see. The latter approach must have developed a bad reputation for introducing as many errors as it averted.
I have seen many a census entry where I have had to scan other names on the page to 'decode' the enumerator's handwriting style and confirm it is the person I wanted.
I can only conclude that the transcription work was farmed out to some subcontractor where the workers are under pressure to transcribe X-hundred census lines per day and do not have the luxury of spending 3-5 mins staring at one name.
With regard to nicked ancestors, my current pet theory is that they are interrogating the IGI and, like me, drawing a complete blank when they look in the right village and era. They widen the search in space and time until they finally get a name match. Almost without realising it, they have leaped to the conclusion that the IGI is fully populated with data.
I suspect it isn't. If your ancestor is searchable online, it's only because of the (dreaded) "user-submitted data". Some other (American, LDS-type) person's properly done research led them to the correct parish microfilm record, which they've paid for a copy of and then typed their findings into the computer system, in a voluntary capacity.
They have no reason to type in thousands of other parish records, for people nothing to do with their family line.
Look how long the freeREG project is taking to build up data. (It's all volunteer transcribers).
And what if the Church of LDS team who came here to microfilm our records missed a register here or there? Did they visit every church in the Isles or just the main public record offices?