b-nodes in rule conclusions

Here's a test (motivated by a current discussion in RIF-WG):

   <a> <b> <c>.
   <a> <b> <d>.
   { <a> <b> ?x } => { _:x <b_inferred> ?x. }.
   _:x <b_given> <c>.
   _:x <b_given> <d>.

The (relevant) output I get from cwm is:
   
   [ <b_given> <c>, <d> ].
   [ <b_inferred> <c> ].
   [ <b_inferred> <d> ].

which makes it clear that the "_:x" inside the rule conclusion does not
name the same thing as the other "_:x"'s in the file do.  In other
words, b-nodes in rule conclusions get an implicit existential
quantifier inserted.  They don't just use the implicit one around the
file.

Two questions: (1) is that as it should be?
               (2) how strongly do you feel that way?   (would it
                   be reasonable to do it the other way, perhaps?)

     -- Sandro