Every time I have designed continuous beams in RAM Element, I have modelled them with a single member across multiple nodes. When I compare the results to a beam modelled with individual members between each node, the results are similar, but not exactly the same.
Which way is the more optimal way to model continuous beams? Was the way I was using it okay?
You'll get the most exact analysis results if you model the beam in segments, but the difference is usually small. The choice is usually based on the design criteria. Things like unbraced length which defaults to the defined member length. My default approach is to model a beam that supports other beams as a physical member (i.e. a girder), but for a cantilever or multi-span beam that is supported at multiple locations, I prefer to segment.
Is there a reason why the results are different?
It's the same for most any finite element application. The program starts with nodal results and then derives beam results for periodic stations along the length. The derived results are never as precise as the nodal results.