It looks to me like it was scanned from print. So at a guess I wonder if this has to do with the scanner compressing the pages with JBIG2’s lossy mode and its “eh, close enough” approach to glyph tables. That is, the approach that swapped 6’s for 8’s on construction documents, among other examples.
Discussion here at the time:
It looks old enough that it probably wasn't OCRed and if it were, you'd think a journal would fix the typos.