For every new schema it is parsing, the implementation looks for a fully qualified schema name, that is already stored in the registry and is doing string replacement in a given schema.
If the resolved context variable content is null, no replacement occurs and the variable name remains unchanged in the query string (including the "%" signs).