I've been using Beersmith 2 for ~3 years now for formulating recipes, and periodically using a hydrometer for measuring OG (mostly on new batches). I also calibrated the refrac with distilled H2O per the manufacturer's directions.
I use a default of 80% for my total efficiency, which has matched up when I had taken OG readings w/ the hydrometer for prior batches.
On my last 2 batches, I noticed the following changes in efficiency:
Bells 2-Hearted clone: predicted OG: 1.070, measured OG: 1.075. (84% measured efficiency).
Saison (brewed today): predicted OG: 1.060, measured OG: 1.068. (90% measured effiency).
I did nothing different process-wise than I've done in the past; same batch size, grain bills, etc. I guess it's time to retire the hydrometer for good, and probably dial down on my grain bills a bit!