PRO:1. Of course, it is not always difficult to understand whether a varia lectio is a typo or whether it has a certain significance. A seeming typo could be only what is left of an older varia lectio.
2. mantrabhāva is certainly a significant variant of mantrabhāga, so why should not also a similar shift of a single phoneme be recorded in all cases?
3. In case of a particularly significant manuscript, even typos may be helpful in throwing light on its background. One could learn, e.g., which kind of typos were more current among scribes coming out of a less known scribal tradition.
4. Moreover, as suggested to me by Péter-Dániel Szántó, one could choose to record all variants so that a future reader of the manuscript(s) we have been using may use our critical edition in order to learn the manuscript(s)' script.
i) In cases such as the ones hinted at in 3), the faithful reproduction of the entire manuscript may be the most suitable solution. So, one could produce a diplomatic edition of the manuscript and then avoid including typos in the critical one.
ii) A too heavy critical apparatus makes the text less readable and is, hence, a non-sense in case of previously unedited texts.
iii) Too many typos in the apparatus make the significant variants loose their significance. They are, so to say, overwhelmed by typos.
iv) Cladistic tools are not reliable in producing a stemma codicum if they have to take into account hundreds of typos (which should play hardly a role in determining a stemma).
More in general, the point is: what does one aim at, through one's critical edition?