However we must take into account that 

The SBJ file from that webpage must be the "expanded" version created ~1400 CE.  The version that would be available to the Author is some "original" version, which was claimed to have 365 paragraphs.  So the parag count of that "original" version is still consistent with that of the SPS, but the word count may not match. That "original" Chinese version is now lost, but may survive in other languages.
In the SBJ file, each word (as produced by Google Translate) may be one or two Chinese characters. (I didn't see any three-character words, but may have missed.)  On a visual check there seem to be ~6 two-character words per line; so the number of Chinese characters should be about 10930+6*370 = ~13100
On the other hand, when creating the SPS file, both the word gap marker '.' and the possible word gap marker ',' were mapped to '.', so the number of words (as intended by the Author) may be somewhat less than 10474.
On the third hand, it is possible that the Reader pronounced each Chinese word, not each character, as a single word.  If that is the case, the 10474 word count of the SPS would be consistent with that of the SBJ above.
On the fourth hand, some of the candidate languages have been described as being "sesquisyllabic" in that some or most words consist of a "weak" unstressed syllable and a "strong" stressed syllable, the former having a simpler structure than the latter.  If the SPS is a transcription of such a language, it is possible that the Author wrote each of those two-syllable words as one Voynichese word.


More to follow...

[/quote]