+1 to all of Natalie's feedback. That's great advice. =)
Given the participants are all experts, you'll probably want to be very clear in your expectations for the purpose of the discussions. If the GameStorming book works for what you're after, it will be excellent in framing the activities. In addition to being a very helpful book, the authors are lovely people. Here's the blog (moderated by Dave Gray) that goes with the book: http://gamestorming.com/
Having more than one group lets you split people into different groups if they work at the same company, A/B test the stimulus material you give the groups, get more/varied feedback, and also swap the feedback from each group to the other to build on it. If it was all just one big group, some of the details wouldn't have a chance to surface. A balance between:
o Being somewhat homogeneous - if people have similar opinions/experience/perspective, they feel less judged and are more able to open up and share
o Being somewhat diverse - if people are too similar then you're not going to get enough breadth of discussion
Expert panels also do better as smaller groups, so if you did get 6 or 7 participants, you could have two groups of 3 or 4. Too many cooks spoil the broth. =)
Record all the audio for later, and have a go with recording the discussion visually on a whiteboard as you go too, so participants have stuff to refer back to as they talk.
And yep, managing dynamics of the group is pretty key! To make sure the more cautious or less vocal members are not only able to participate, but also are given space to contribute.
For expert panels, the only other thing I'd add is to make sure you stay focussed on information generation, not problem solving (which is easy to devolve into when you get a group of experts in a room and they get a whiff of the problem space you're working on).