Hello Clay,
Not sure if this is normal or not - probably, a better place to find this out is one of the FastMM support resources:
http://sourceforge.net/projects/fastmm/supportFrom the profiling perspective, if the total execution time of this function is just a small fraction of the relevant operation's execution time, you should probably ignore this one for the moment. Anyway, even if you manage to reduce the number of calls, the maximum time you win can't be more that 0.32 seconds. Is it worth while spending your time on this? You could probably start the optimization from other points where you can win more time with less efforts.
Good Luck with the optimization!