What you are saying is that the meaning is arbitrarily assigned by you. If you do this for any reasonably ordered text you will get similar repeated words and your brain will fill in the meaning based on your desire to find it. The method will _not_ fail along the line even if you are completely off from what the writer "intended". In fact, you are more likely than not to find meaning if the text already has a medium to low entropy such as the VMs content does.
I can try to write a simple demonstration of this using randomly generated text based on the voynich statistics. I'll try to do that over the weekend.