It was originally a '5 group' script but the problem shows up with four groups as well (MZm, DZm, MZf, DZf). When I fit to four-groups (first run in the job) the likelihood is greater (12034.08 instead of 12030.299) than the sum of the likelihoods of the two two-group jobs (second (males only, 4973.186) and third (females only, 7057.113) run), when it should be just the sum as there are no constraints across the male and female groups. When I check the sum of the likelihoods in the individual groups, the likelihoods for females are exactly the same in the 2- vs 4-group job, but the likelihoods for the males are not. The problem gets worse when adding the DZmf group. The precision of the jobs is the same. Classic Mx gives right likelihood.