When I first worked for the Bodleian, in the mid-1990s, I became aware of an issue that still causes trouble.
If, hypothetically, the Bodleian had acquired a manuscript on the third day of a five-day sale which began on 1 January, they recorded it as having been acquired on 3 January. Of course this is completely accurate, and represents a historical fact. A problem arises, however, because most people would cite the auction only by its start-date, in this case 1 January, which creates an apparent contradiction. It can then cause a reader some trouble, until s/he realises that the "1 January" catalogue and the "3 January" acquisition both refer to the same auction.
What I realised clearly for the first time is that most people conflate two very different and separate things: on one hand there is an event in the past, the auction, and on the other hand here is a bibliographical entity, the auction catalogue, that may not represent the historical event with complete accuracy.