Im not 100% sure but as far as I know there is really no evidence to support this "dangerous" behavior from the so called unpredictable stalls. No, the five digit series of sections are not exactly optimal but for a "turbulent" section family they do have good l/d numbers and have a low moment coefficient as compared to similar counterparts.Early on, I was enamored of the 5 digit ones for their low Cm, but can not now stomach the sharp stalls; too many
pilots have turned their last finals with these (hint).
True, the NACA data suggests that the sections may exhibit abrupt boundary layer separation but the wide practical application of the shape and decades of history suggests otherwise. The family has been used by Taylorcraft, Cessna (all the way up to the Citation), Douglas (DC-4 to the DC-7), Dornier, Embraer, Helio, Piel Emeraude, and on and on. And in none of those cases have there been any documented cases of deadly or for that matter, even dangerous or unpredictable behavior (at least in 30 years I haven't seen anything conclusive in that direction).