Why dont you attach a sample file with enough samples for us to identify the variations in your data?