Indexes, Nulls And A Join Problem

I am getting VERY poor performance on a join even though I have indexes. Instead of using the indexes, SQL Server does a table scan on one of the tables in the join. After I force the index in the query, I can see why. The index doesn't give any better results than a table scan. I suspect it may be because the table has a huge number of rows with null values for one column of a two-column key used in the join. However, the same tables use the indexes and give good performance when queried independently. It seems the problem is specific to join processing. I notice that the index I created appears to have pages referencing the null rows. I know that Oracle disregards nulls when creating indexes, but does SQL Server not do so? Is there any way to create a better index? And if I can't solve the join problem, any suggestions about workarounds?

The details:
-I have TABLE A with columns X, Y, and Z. In TABLE A there are 380,000 rows, and X and Y yield a unique key. The index is on X and Y. This table has 240,000 rows where X is null but there is a non-null value for Y.
-I have TABLE B (60,000 rows) with columns X,Y, and Q, where X,Y, and Q yield the unique key and have an index.
-Here is my query:
SELECT A.Z
FROM A, B
WHERE B.Q=1
AND B.X = A.X AND B.Y = A.Y