To me, this looks to be a sign problem in the FRED model.  The variance in 
all cases is exactly double the flashed amount.  Jim please verify if this is 
correct.  Thanks.
PL