I think they were as there is an earlier microfiche version of IGI well before it went online which I'm sure had more extracted entries (can't easily check until next week) - i've sometimes followed up missing entries found from actual register - there would appear to be two distinct types of batches 1: appears to be wholesale corrections to a single register (St Marks is a good example in that extracted entries are often misreads whereas certain batch entries are accurate), Michael is somewhat different as the film of the official transcription contains 2 overlapping registers as the Browns (father + son) transcribed part of the old register prior to new entries - this transcription contains errors when compared against 'original' earlier reg.
2: the more usual (and common case for marriages) is that a complete family name (Corlett is a prime example but each parish will have one or two fouled up names) has been 'replaced' by patron submitted entries - sometimes you can determine the tree that was submitted as a certain batch contains several related entries from different parishes that are related.

The problem is that most of the early extracted records are very accurate - since they were working from the same late 19th C hand this should be the case - possible double keyboarding was used to check for errors (a common technique for entering data in the days of punched cards) but later batches indicate many more misread entries (jan + jun confused etc)