Actually Luis, yours gives the closest to the needed result on the actual data. With some tuning I think it's going to work for me. Thanks to everyone for their help!