In a third run with the fluorine torch, the settling chamber wells were kept hotter than before (≥ 625°C); the flame was cooled by diluting the fluorine with helium. In the analysis of the products, 99% of the thorium fluoride fed in was accounted for, but only 64% of the protactinium activity. Part of this was carried in the exhaust gases past the cold trap and into the soda line disposal column, where it was detected by survey meters. The stripping of protactinium from the solid was somewhat more efficient than before; 77% of the feed which was recovered from the settling chamber had lost 72% of its original specific activity. About 15% of the input activity was trapped on the cold fingers with very little thorium fluoride.