Well, the OCR read the documents but most of the documents had some sort of stray marks placed there by the respondents.
OCR technology is now so advanced that the USPS can read virtually 100% of all handwritten addresses, or printed addresses of any font.
I'm thinking the OCR used to originally "scan" the source documents for input into the microfiche system was a low res standard TV camera. So there'd be all sorts of "noise" in the signal that a modern software with a gazillion pictel resolution could probably pick up and toss into separate "layers" or "buckets" or whatever we wanted.
I know most folks don't think of TV as being just another OCR since it's analog, but there you have it. If all you wanted was to spit out a downstream "photostat" you'd simply record the full analog image. Digitization of that image definitely has to be closer to 2000 than to 1961 I'll guarantee.
I got Abobe Acrobat 1 with an OCR app in a scanner bundle in ‘96. The scanner was new but the software by that time was already a few years old. What a PITA that stuff was to use, and the results were horrible.
You are right, the process has come a very long way. Think about all those hand written checks running down the line at the federal reserve.
I ran the AP version of the BC through a little freeware app called PDF OCR X, for intel chipped Macs. It came out nearly completely converted. Each layer was labeled with corresponding text. Pretty impressive considering the resolution of that doc.
IIRC HDOH converted their documents in 2001? So we aren’t exactly talking about a stone tablet here.