Are perfect-reduplication and quasi-reduplication related?

[MarcoP]

Before tackling page (and/or folio) stats, I experimented with the scrambling idea suggested by Koen.
I attach a graph based on the original/scrambled ratio for reduplication (X axis) and quasi-reduplication (Y) counts. For comparison, I added two ordinary texts: King James Genesis (English) and Mattioli (Latin). I also include Timm and Schinner generated text (TT-AS).
For the language texts, the original has a lower rate of reduplication and quasi-reduplication than the scrambled version (orig/scrambled ratio < 1). In the original texts, reduplication and quasi-reduplication are close to 0%.
For the VMS samples (and TT-AS) the ratios are greater than 1: more reduplication and quasi-reduplication in the original then in the scrambled versions.
The BioQ13 section has high reduplication and quasi reduplication percentages, yet the scrambled version is particularly close to the original (both ratios between 1.0 and 1.5). In the case of perfect-reduplication, this can be explained by the low MATTR, but the result for quasi-reduplication is more difficult to explain: I guess it means that many of the words are extremely similar to each other. Overall, the VMS sections display a great variety in the values for these two ratios, but they consistently are >1. I guess this means that there is some principle in how Voynichese select consecutive words that causes reduplication and quasi-reduplication: they are not explained by the nature of the vocabulary alone.
TT-AS behaves quite similarly to Voynichese (Herbal_A in particular): since we know how this text is generated, it could be interesting to see exactly how it manages to replicate this feature.