Christian Specht wrote a short, cute analysis on citation mutations. He has a follow-up. Basically, these result from typos by authors or typesetters. This isn’t the problem. The problem is that some typos are inherited. Specht speculates that the inheritance (i.e. copied and propagated through citations in other papers) is a problem because it implies that authors simply copy old references from other papers. I guess the ideal would be that authors would use their own database references or to build up their citation from the actual paper.
I ‘m not sure if this problem is as distressing as Specht writes, although to be fair he isn’t exactly worried.) He simply made a point that there is likely much copying of old references – even if we can’t detect the occurrence because most people usually copy the correct reference.
Specht worries that the incorrect references may be an indication that scientists do not always read the papers they cite. I would add, simply, that maybe some scientists are lazy; if a paper already contains a properly formatted bibliography for the journal to which a new paper is being submitted, I can see why some authors might simply save time and make a copy.
Or the level of scrutiny for a paper usually doesn’t reach into the bibliography, which, ideally, would involve the authors actually searching for each paper and actually checking if the page numbers match those from the article.