I'm not 100 percent certain of the statement I'm about to make, but, from what I could tell...

The number replacement problem happens on everything, not just OCR. It happens on copies, scans, making PDFs, everything. These devices don't work like old copy machines: everything they do that involves that scanning bed starts with a scan and a conversion into digital form, and THEN the next steps are applied from that digital form: printing it out (ie, making a copy), saving the scan (even if it's just intended to be a bitmap of the original), emailing, PDF conversion, everything.

The number replacement issue happens in the first step, where the image is initially data-compressed before anything else gets done to it. So if you own an affected unit, you're screwed no matter what.

And according to that article, upping the quality of the scan doesn't fix the problem.