What if Voynich plaintext is very heavily shortened?

[Juan_Sali]

It also seems to me quite probable that we shouldn't treat each of Voynich letters independently but rather work
with common groups of letter like or, ol, ain etc.

If we did so, another problem emerges however. If we assume that groups of Voynich letters code single letters in the plaintext
then the supposed words in the plaintext become really short - usually 2 or 3 letters.

If we treat space as space then text made of such words would be generally very improbable, unless we assume that the words are heavily shortened.

The words length need to be hidden, so the spaces in the VMS cant be the real spaces. A way to hide words length is a continuous text without spacing, another way is adding spaces. The second one is worst that the first as it gives more clues of the composition of the n-gramms. 
Why to choose the second one? A possible explanation is to easy the reading for those who know the code, so they can read the VMS directly.