In the past, it's been pretty easy. I have only had one project at a time, so my plan for talks was: 50% present project, 50% most high-profile past project. Add in a dash of 'clumsy attempt to link them' and voila.
I have to give a couple talks in the next few weeks. Now things get more complicated.
Like most parents, I wish my children had more citations, so I feel some need to brag about them in public.
Also, those ideas are the ones that are demonstrably good enough to publish. I know how to present them in an entertaining way. I don't have to worry about the audience member who raises his/her hand and points out the major flaw that makes the present study worthless. Plus I already have the slides made.
OTOH, those come from my earlier life as a post-doc. And I want people to realize I have my own lab with new ideas coming out. I could talk about my current research, even though it's unpublished. This has its own advantages:
(1) Advertise my young lab, and my independent self
(2) Advertise these ideas, which may help in the review process
(3) Attract comment and criticism, which could improve my ideas [very likely in this case]
(4) attract potential collaborators [possible, though unlikely, I suspect]
(5) Advertise my students [which is part of my job as PI after all]
If I could I would write a brilliant talk incorporating all my past research and the new stuff in a graceful and elegant way that put the audience into a state of amazement and wonder. That's a little tricky though. I think I will be focussing on the new things.
There's also the question of focussing solely on the projects that seem to be working, or also mentioning the ones that don't totally make sense yet. And that relates to the question of how much larger framework to give, versus how much to just say, here are some unconnected results.