I was assuming that this applied
http://www.roxio.com/en/support/cdr/howrecworks.html The "annealing" process - a "physical" erase.
And therefore, no deviations to the different response of areas in crystalline or amorphous state.
I would suspect the properties of an overwrite to vary slightly, depending on the initial state - that seems like the only way the erase condition could influence the result.
If the full erase is only LOGICAL, then it's still rewriting over a random condition.
The only other thing I can think of, is if a full erase is performed immediately prior to writing, the media will have had plenty of time to temperature equalise.
Other than the heating effect from the write laser, equalisation would also occur if a "test and burn" was used - THAT might be a way to pin it down....
Full erase immediately before,
Full erased earlier,
Quick erased, and burned with test before.
If my temperature equalisation idea is right, the second one should be the worst?