We know that we need to take special attention about NULL values while writing queries because they may lead us into writing incorrect query logics. NULLs may also affect the performance of aggregation function MAX() and MIN(). I have submitted this issue to Microsoft Connect. There are number of ways to get around it. I think the improvement can also be done at query engine level. Hope this can be fixed in next or future version of SQL Server. But for now, we need to know it and know how to get around it.