David,
Very interesting observation indeed. I don't have the answer, but a possible approach.
I am assuming the following:
a. Heat transfer area of the overhead/reflux condenser in your pilot unit is not limiting.
b. In the case when MIBK is not present in the feed, your pilot unit is operating almost at total reflux. Only up to 1.12% of the feed to the stripping unit forms 25% (w/w) ammonia containing distillate.
You may like to please check from the literature, whether ammonia-water-MIBK or ammonia-MIBK form minimum boiling azeotrope. If they do, then the boiling point(s) of the azeotrope will be lower than the boiling point of ammonia. Water at 40 deg F will not condense such an azeptrope, it will 'escape' the overhead/reflux condenser of the pilot unit.
If this is correct, two stage condensation, as suggested by Milton, may not help. A better bet would be to remove MIBK, perhaps by adsorption, before the stripping section.
Please post your findings from literature search for azeotrope data on this site's message board.