BaseX's behavior is somehow influenced by whether a document uses namespaces or not? Is that true? In other words, BaseX shows different performance depending on whether prefixes are used?

Why is that so? Why did you not choose a representation that abstracts away namespace prefixes and treats all elements as if they were in a namespace? Logically, it's not a higher burden (storing a pair (index, string) is just as efficient as storing or indexing a string); and BaseX is not required to output XML that preserves prefixes and default namespaces when a query is made; any equivalent XML that expresses the query result is acceptable.

I'm wondering if the correctness problems we're seeing with respect to namespaces may be due to this decision of treating namespaces specially rather than the default case.

- Godmar