Been spelunking in the files for personal-submod reasons, and I think some of the weighting modifiers for the effects when a Lollard or Waldensian attempts to translate the bible are backwards. The negative outcome has modifiers with weight 0.8 that trigger when the character has learning below various thresholds, meaning that whilst a character with learning 9, 10 or 11 has equal likelihood of getting the good or bad outcome, a character with learning below 4 has about a 66% chance of getting the good outcome, a higher chance than any character with learning below the final threshold of 20 (for comparison, a learning stat of 20+ gives a chance of ~68%). Switching lines 1325-1327 with 1328-1339 in emf_heresy_decisions.txt would appear to fix the problem.