One Glyph-multiple possible readings- coder's tool?

[BessAgritianin]

I also assume that the boundaries between the "letters" in Voynichese are blurred. In theory, the intended word should be guessed from the context, and there should be no problems.

Thank You for this observation. It alighns with my hypothesis. But a system which uses context to resolve ambiguity cannot be read mechanically and cannot be decoded using substitution on frequency analysisi alone.
  Let accept that the next word is "woes". This means that we are dealing with some kind of affliction. But here is the problem- "roud" coud be read as stone (German- stein) (kidney stone), could be tooth (German -zähne) as toothache, coud be birth (Czech-rod) as birth pains, could be retin (Latin) for retina problems.  The coder obviously (from the marginalia readings) uses Latin, German and old Czech.
 He is probably medical person who knows several languages. Therefore a person, who aims at interpretation the text needs also to know what is the plant and what are its properties for healing... In many cases we do not know the plants...
Therefore the coder system rewards the knowledge and not the mechanical AI translations, like we are used to do.