OCR scans of letters is not the issue. When two boxes on the same single (supposedly) scanned document are identical, as if the original printer of the form made absolutely exact separate boxes, that is evidence that a box was picked up and used/moved via a Adobe/photoshop type technique. But you’re doing a great job running interference, dissembling, obfuscating for the won.
A box, in terms of character recognition, is no different from a letter or number, because it too is just a standard Unicode character. You can even type it □.
To be evidence that it was intentionally copied, it would need to be something that wouldn't be recognized as a Unicode character, like, say, a drawing.