This is strange as the passband insertion loss should be fairly constant. At higher frequency the attenuation should increase, not decrease. Do you have any ripple in the passband?
Could it come from the resonant circuit at the input of the filter (L4)?
Jean-Claude
