@JoJo_Jost, to answer your question 
the transliterations are definitely full of errors.
These errors are also of several different types. Going from the bottom up.
1) There are bound to be plenty of errors in the actual handwritten text on the pages of the MS. In this, I fully agree with Stolfi. We have no way of quantifying this now. If a solution is ever found, we will know then.
2) The transliterations include an interpretation of what are the same characters and which are different. Since different transcribers used different assumprions, the various files are different in that respect. The STA versions of the transliteration files tend to include all differences that have been noted. (Though not the RF file which eliminates a few of the v101-based differences).
3) Finally, people have 'read' the MS differently, so there are transliteration mistakes as well.
The impact of errors 2 and 3 can be avoided by using the original images of the MS, but of course this only allows you to do manual work, no statistics.
It can also be partly quantified by doing statistics or experiments with different transliteration files in parallel.
An error of type 2 that will not be easy to catch is, for example, if in reality f and p are one and the same. Since these characters are relatively rare, this will not be a major problem, but worse cases could easily be imagined, and may well exist.
Finally a word of caution. If a theory is not working, it may be tempting to blame the transliterations. One should always be realistic in which types of errors one is likely to accept, and which should not be acceptable.
the transliterations are definitely full of errors.These errors are also of several different types. Going from the bottom up.
1) There are bound to be plenty of errors in the actual handwritten text on the pages of the MS. In this, I fully agree with Stolfi. We have no way of quantifying this now. If a solution is ever found, we will know then.
2) The transliterations include an interpretation of what are the same characters and which are different. Since different transcribers used different assumprions, the various files are different in that respect. The STA versions of the transliteration files tend to include all differences that have been noted. (Though not the RF file which eliminates a few of the v101-based differences).
3) Finally, people have 'read' the MS differently, so there are transliteration mistakes as well.
The impact of errors 2 and 3 can be avoided by using the original images of the MS, but of course this only allows you to do manual work, no statistics.
It can also be partly quantified by doing statistics or experiments with different transliteration files in parallel.
An error of type 2 that will not be easy to catch is, for example, if in reality f and p are one and the same. Since these characters are relatively rare, this will not be a major problem, but worse cases could easily be imagined, and may well exist.
Finally a word of caution. If a theory is not working, it may be tempting to blame the transliterations. One should always be realistic in which types of errors one is likely to accept, and which should not be acceptable.
![[+]](https://www.voynich.ninja/images/collapse_collapsed.png)