The important point to remember in my opinion is that it is always assumed that the well punched digit is the intended one, however, it is not guaranteed that the intended correction is punched in greater relief than the original. A clear case of this is the GEOE shilling. Nobody is going to change the final R to an E. I believe that the GEOE was punched in, but the intended correction wasn't deep enough giving the appearance of E/R. This is quite easy to justify if the die has been hardened. I can certainly show an example of a hammered coin where the overmark struggled to reach the same depth using numerous blows and there is no reason to expect a die for the milled coinage to be any different.
If the above is added to the list of permutations of die sinkers' errors and corrections, you are led to conclude that virtually everything is possible. e.g. Sometimes the date starts too far to the left or right, the correction leading to apparent overdates.