Ah, OK , right - Sweep2 does not split at tangents - no setting for that - I see that Sweep1 does split, because the default behavior is not to refit the rails, whereas Sweep2 does refit the rails. My guess is that Sweep1 can make simpler objects that reflect the curve structure better simply because it does not have to contend with another rail - it is rare that they are the exact same like this in Sweep2, usually the surface must be cooked up with different inputs on each side. If SimpleSweep is not available or not used in Sweep2, then the rails are refit to G2 across G1 joins, I believe.
CreaseSplitting should in general always be set to Yes - it ensures, if it’s working, that surfaces with actual creases (G0) are split up. It does not split ‘potential’ creases however. In Sweep2, if you use SimpleSweep, the rails are not refit and you can DivideAlongCreases the result. CreaseSplitting determines whether surfaces created behind the scenes are sent through the ‘kink splitter’ to be made into polysurfaces before being added to the document - it has no provision that I know of for dealing with G1 continuity, only actual kinks.