Large Object Support is facilitated by removing some of the index information from DB2, and putting it into a header that's stored with the object itself. I've run into this before while trying to estimate the volume of storage required to do a data extraction. I'm not aware of any published specification for that header, so there's no quick and easy solution.
There ARE ways of getting this data if it's absolutely critical -- but your best bet going forward is to enable page counting in one of the more recent versions of CMOD, so that the number of pages is a value in a DB2 field.
-JD.