That is true, that is why I always say it pays to be careful to look at things in the 6z and 18z runs-there is no new upper air data to be input at that time. The off-time runs may best be described as updating from the on-time runs.
In winter storm forecasting, you will often read the writer of the discussion say he wants to wait for the 12z or 0z before making big changes to a forecast.
There are SOME times, however, that those off-runs do pick up on something and start a trend that verifies. The most recent example is with Ivan. I believe it was the 18z GFS that showed Ivan getting picked up by the trof and not being stuck in one spot.