Well, the number of morse code signals needed to encode a single letter ranges from 1 (for the most common letters in English such as "e" and "t") to 5 for the less common ones.  Let's say it averages out to about 2.3.  Then there are also the nulls, which occur fairly frequently (anything not strictly a curve or line glyph, such as the VMS "d" or the gallows characters, is used as a null).  So, add on another glyph-per-letter to roughly account for there being, on average, about one null per "word" (of course, in my system, each "word" separated by spaces corresponds to a complete morse code letter).  So, I'd guess it takes about 3.3 Voynich glyphs to encode one cleartext letter using my system.  

Regarding whether my system produces text similar to the VMS, the important distinction is that it CAN, but it does not automatically do so, as the encipherer has many degrees of freedom for how to encipher a particular letter into obfuscated Morse Code using the Voynich glyphs.  Basically, a dot just needs a curve glyph of some sort, and a dash just needs a line glyph of some sort.  So, if you are determined to make your code "look like" the VMS, it is quite possible.  

However, by itself this does not explain, for example, why THE ORIGINAL VOYNICH ENCIPHERER would, for example, prefer to have nulls (gallows) in certain places in vords or lines, or why certain glyphs would tend to appear towards the ends of lines (i.e. why THOSE particular curve or line glyphs, and not some other suitable curve or line glyph?)  The only way to assert that this (or some other system very similar to this) WAS the exact system that the original voynich encipherer used would be to assert that this encipherer held to these extraneous patterns (i.e. extra patterns that didn't carry semantic meaning and that thus were open to many degrees of freedom) for purely AESTHETIC purposes—i.e., that the encipherer just "liked the look" of gallows characters in certain positions in vords and lines, or that he/she just "liked the look" of lines ending in certain curve/line glyphs and not others, etc.  

Possible reasons for why an encipherer might "like the look" of certain non-semantic patterns over others might include the desire to make the text superficially look like it had Latin abbreviations or pilcrows, etc.  But these suppositions are shaky ground, so at this point I would not claim any real likelihood that the original VMS encipherer used a system similar to mine.  The point of my exercise was simply to point out that this, among others, was one possible way it could have been done.