But I have a question about some behavior I am seeing in the following-sibling axis that does not seem logical to me.

Here is the data stored in basex:

> xquery /text[@id='test']

<text id="test">

  <clause>

    <word>A</word>a

    <word>B</word>

    <word>C</word>c

  </clause>

</text>

Here is the query on that data:

> xquery /text[@id='test']//word/concat(text(), ' ', normalize-space(./following-sibling::text()[1]))

A a

B c

C c

I think the result you get is caused by whitespace chopping during XML parsing, seems to be the default, see http://docs.basex.org/wiki/Command-Line_Options