We used West and Kenny’s (2011) Truth-and-Bias (T&B) model to examine how accurately group therapists’ judge their group members’ alliances, and the effects of therapist-patient congruence in alliance ratings on patient outcomes. Were considered: (a) directional bias – therapists’ tendency to over- or underrate their clients’ alliances, (b) truth strength – clients’ alliance ratings, and (c) bias strength – therapists’ tendency to conflate their alliance ratings for a specific group member with the average alliance ratings for the other members of the group. There were 118 obese adult patients with binge-eating disorder that were treated by 8 therapists with Emotionally Focused Group Therapy. Outcomes were operationalized as pre- to postchanges in: health-related quality of life, binge eating, and psychological distress. Patients’ and therapists’ working alliance were assessed after the 2nd, 10th, and last (20th) group therapy sessions. (a) There was no significant congruence between group therapists’ and members’ ratings of alliance; (b) therapists’ ratings of an individual group member’s alliance were significantly related to therapists’ ratings of the other group members’ alliance in early sessions but unrelated in later sessions; and (c) the relationship between therapists’ alliance ratings and bias strength was weaker when patient binge eating outcomes improved. Group therapists adopted a “better safe than sorry” strategy by underestimating the strength of their group members’ alliances. Therapists had a tendency to judge each group member’s individual alliance based on the aggregated alliance of the other group members. Improvement in patient binge eating outcomes was related to therapists overcoming this tendency.