As a research scientist, I can not tell you how many scientists seem to have watched a few TED talks too many and desperately try to be interesting, never actually telling me the essence of what they've done instead of some analogy or a pretty picture.

That being said, I don't disagree with you. The point is to know your audience and your goals. Trying to be interesting is great for a lay audience, as are long derivations for a seminar of your peers. Just don't confuse the two.

I think the best way is 3-5 word bullet points, but mostly talking about tables/charts/figures. If you've truly done the research yourself, you can easily explain what is behind every chart.

Really depends on the content. If I'm being presented with a mathematical theorem, for instance, I'd like to read the exact statement at my leisure while the speaker reads it and then talks about it. Structuring derivations can also be tricky (omit too many technical details and risk losing the interesting part of the work). Charts are fine, but for an audience of scientists the methodology behind the result is often more interesting than the result.