Hi Anil,
I think my next step would be to break the activity down into component parts. Currently you rely on rebuilding the clustered index to rebuild the nonclustered indexes. I would break that apart into individual commands and run sybmon on each to see if the unpins are specific to any one (or if the unpins only occur when rebuilding the clustered index is also rebuilding the nonclustereds)
drop indexes
create clustered index
create nonclustered index 1
create nonclustered index 2
I can also use the additional information you've provided about the table and indexes to see if I can cause unpins on my test systems. A bit more might help, could you share the index defintions? How large is the data cache used for this table?
-bret