SQL is ignoring my index hint and using a clustered index.

Part of this gets information from a table where an index hint has been included.
Looking at the execution plan, SQL is ignoring the index hint and choosing to use the clustered index on the table, which is not efficient.
The fragmentation on the table <=8%. I have come across this before and rebuilding the indexes does work, but this will require downtime as the table is in constant use and SQL is not enterpirse edition. Is there any way to force SQL to use the index hint provided to it?
I know that when I do a reorg/rebuild the query will pickup the correct index, but I would have thought that the hint would have been enough to make it do this without having to do anything to the stats.
Any general information on this would be useful as I am not having issues with this particular proc, I have just seen this before and wanted to know why SQL does what it does, not the 'fix' for it.
Thanks
I'll have a read up on these! I agree that the hints should only be used as a last resort, but it still seems odd that hints can be ignored in the execution plan for a more inefficient index.
Thanks for your help,
Dali
The db is SQL2005 and I don't believe filtered indexes are available until 2008.
Sorry I didn't mention that before.
Ed
Re-factoring the proc is the best solution but, as previously mentioned, it is just interesting to understand why SQL ignores hints.
In this particular case, I re-organised the index which resolved the issue. However, as the fragmentation of the index was only around the 8% mark, it is interesting that SQL decided to ignore the index hint and use another index of similar levels of fragmentation (0.03% difference in fragmentation between indexes).
In general, index hints should not be used, so looking at the actual query will be helpful in understanding what it is attempting to do to cause it to not use the index. One of the more common reasons that an index is not used is: Statistics are out of date, so update those. If the NC doesn't cover the required values (either through the indexed fields or with the includes) and the number of seeks + key lookups > scan, then the clustered is often used. Also, although not as good as a rebuild, you can do an index organize. It is an online operation that can help in such situations.
Ed