No, beam theory falls apart if the beam is too short. The equations no longer hold true.
You can simplify structural beams to optimize performance in Simulation to be modeled with beam elements. The size of the problem and the resources required are dramatically reduced in this case. For the beam formulation to produce acceptable results, the length of the beam should be 10 times larger than the largest dimension of its cross section.
You can see this explanation on SW help. Check this elements for this condition.
I disagree; results may be fine. They may also be off.
The warning message pops up if the L/D ratio goes below 10.
I like your question. Splitting the beam into two pieces does seem 'less bad' than simply using a short segment within a larger structure of beams.
If you go to help -> sim -> validation -> verification problems you'll see "bending of a solid beam" halfway down. Run the 'force beam' study and you'll see the results are .0051 in displacement, vs .0050 calculated. That's a 2% error. If you split the beam in half, recalculate joints, and rerun you're now at .00512 in displacement, which represents an additional .4% error. That's pretty benign.
You can also put a split line on the weldment beams, and apply loads to the split line.
Validity of results depends on your tolerance of errors. Are you just optimizing, or checking hand calcs, or is this the sole calculation that determines if the structure collapses and a person dies?
Guess it all gets back to allowable error. Some cases may be able to live with, others not so much.
Your comment mentioned using a split line on weldment beams. I tried this before but can only select vertices, joints, or the entire beam. Can not select either face about split line or the split line itself. This would be solution if worked.