Diagnostically-homogenous groups have some advantages since this makes it easier to tailor groups to participants' needs. However, one can also do a general CBT group effectively as well. One approach is to present general principles and widely appropriate techniques (such as thought sheets, activity scheduling, in-vivo exposure, etc.) and then explore how they apply to each group member's issues. This takes more thought and effort on the therapist's part than a group where all participants are dealing with the same problems, but can work well.

A recent post to the ACBT mailinglist recommended having a co-therapist who is also skilled in CBT rather than having a single therapist lead the group. This makes it easier to respond promptly to each participant's questions and concerns. Also, one therapist can attend to group process while the other is presenting content.

One diagnosis where number of experts recommend a disorder-specific group is Borderline Personality Disorder. Reportedly, having one or two participants with BPD in a heterogenous group tends to disrupt the group while a group where all participants have BPD goes better. (I'd definitely want a co-therapist) Marsha Linehan's manual for DBT groups would be a good starting place whether you're doing DBT or CBT in general.

Mind Over Mood should work well for the kind of groups you've described. In addition, David Barlow has written a number of treatment manuals that may be useful. They can be found at http://www.bu.edu/anxiety/dhb/treatmentmanuals.shtml While I can't vouch for it personally, I've heard good things about Cognitive Therapy in Groups: Guidelines and Resources for Practice by Michael L. Free (Wiley, 1999). Reportedly it proposes session-by-session protocols, includes handouts for clients, and even has the overhead projector illustrations that you can use to help communicate the theory and practice effectively.

I don't do group therapy myself. I hope forum participants who do will contribute their observations and suggestions.