Tombstones are not being cleared up

We deleted part of data from one of column families. After that we ran repair, followed by reducing gc_grace_seconds to have tombstones space to be reclaimed. After this, we kept the compaction enabled to reclaim this space upon passing gc_grace_seconds. We have done this while back and for sure gc_grace_seconds has elapsed.

Currently, when we run queries against this column family, we are seeing lot of tombstone errors like below:
WARN [ReadStage:1113] 2015-05-06 17:48:22,556 SliceQueryFilter.java (line 231) Read 19 live and 1140 tombstoned cells in YYY (see tombstone_warn_threshold). 10001 columns was requested, slices=[-], delInfo=

The localDeletion reported above is time in future (year:2038, I believe it is the default Integer.MAX_VALUE), and I believe this could be the reason it is not being reclaimed.

May I know, how do we purge this tombstones and also, what could have set this date to time in future?