Finally Quick update,
My assumption helped me little.
I ran dwain's sugestion on my production box and it ran in 10 sec.
Tried my logic as well and it took 11 sec.
The reason i got it 28 second is i was running the query using Virtual Path Network. thanks dwain for your help.
Any more suggestion is also welcome. Would like to leave this thread open as i am hunger to reduce the execution time.