Are you using Speakerworkshop? If so, did you measure the pulse response and and then set your time window to exclude the reflections that showed up? It looks like you did a gated measurement so you need to recognize that the result of that is only really 'valid' down to a certain frequency and the low end will kind of be smoothed out. That big bump around 1khz might really be lots of little peaks... If you're using SW (it's just what I'm familiar with), you might try doing a measurement without gating, using 1/16th octave smoothing on it, and then looking at how that compares to what you're seeing with your current measurements. I usually get pretty good correlation between types of measurements when I have a decent measuring setup.

Also, how far away from the mouth of the horn was your mic? You can measure some weird things when the mic is too close... I would suggest around 1 meter away for normal size high frequency horns.
Setting the horn on the floor might not be the best setup, depending on the size of the horn. I usually just set the horn where I'm going to put it to listen to it. Obviously this won't work well if you set it right next to a wall, but then again, that will usually screw up the sound anyway.

John