In-flight reported delta-V for a stage with vacuum optimized engine in atmosphere (high altitude) is much worse than as reported by the designer's performance analyser at same altitude.
In-flight switch-over to much higher delta-V happens at 54 to 58 km (Droo), probably depending on the exact configuration of the engine.
revision:
The issue is not so much that in-atmosphere dV is much lower than designer reported dV in vacuum, but rather that dV 'jumps' up to much higher dV at high altitude (varying from ~53 to ~58km, seemingly depending on craft and ascent profile) instead of gradually increasing as atmosphere density decreases.
Since we need to have a perfect vacuum when space starts so crafts don't decay and can warp without issues there's a region of linear interpolation down to 0 density at the edge of the atmosphere. This is what causes these discrepancies.
It's not the most realistic solution but it was necessary to fix other bugs that came because of having a realistic drag decay.