With regards to the problem you had with the 2 nS vs 3 nS expected rise time on the Rigol, would this have to do with the 1GSa/ps limit on the Rigol introducing up to 1 nS of error? Was this with real or equivalent time sampling turned on?
the error on Rigol E is probably coming from the DSO internal attenuator. While testing the bw with Marconi RF generator
the internal attenuator was off (200mv/DIV), but while testing the bw with pulse generator the attenuator was on (2V/DIV).
From my experience by default all DSO having this particular frontend (Rigol E, ATTEN, Siglent, Tekway/Hantek, etc.)
are calibrated to overshoot a bit while the internal attenuator is enabled, this helps out with the large signals response.
Due this overshoot the resulting bw (calculated back from rise time) seems to be "higher", which of course is not the real case.