There was a high level of inter-rater agreement between the three raters for the NM reports (r = .70) as well as for the M reports (r = .77), indicating that there are systematic patterns in the verbal reports that corresponds to certain positions on the rating scale for both NM and M trials. Even more interestingly, there was a high correlation between the raters estimate and the original rating of the participants for NM (r = .59) as well as for M reports (r = .71), which indicates that the verbal reports in the M trials do in fact track the participants rated level of agreement with the opposite of the initial moral principle or issue [emphasis added] (for an illustration of this process and example reports, see figure S1, Supporting Online Material). In addition, this relationship highlights the logic of the attitude reversal, in that more modest positions result in verbal reports expressing arguments appropriate for the same region on the mirror side of the scale. And while extreme reversals more often are detected, the remaining non-detected trials also create stronger and more dramatic confabulations for the opposite position.

Am I misreading this, or does it say that the verbal statements of people supporting an inverted opinion fit that opinion better than those describing their genuine opinion?