About querying idref attributes

Hi,
  
   I am now studying how to evaluating the query on the XML graph efficiently. One of the possible attributes that make the XML document modeled as a graph is ID/IDRef attribute. It seems that most benchmark queries (i.e. those of XMark) tend to indicate the IDRef attribute explicitly in the query. What I want to ask is that whether it is possible that querying the IDRef attribute is implied?
   Let me explain by an example. Given an XML document about the movies. Every Movie element may have many Actor subelements. Thus when model the document as a graph, there are edges between Movie and Actor elements. It is possible that an actor, e.g A,  plays in more than one movies, e.g. M1, M2. And suppose the detail information of A is stored as an subelement of M1. Than for M2, it refers to A by ID/IDRef attributes. Then when the query wants to returns all the  movies in which A plays, the result should include M1 and M2, right?   But if the query is stated in this way " Movie/[Actor='A']",  will it return M1 and M2, or just M1?
   Thank you very much in advance. 

Regards,
Fianny