I wear my Dornbluth every day, seven days(and nights) a week, so in testing accuracy, for me, the only parameter that matters is how the watch perform on the wrist. As for the reference timekeeper, I use the internet GMT time (see the GMT official site), which they say is accurate within 1 second. I know that there are other scientific methods for testing accuracy of a watch... but, as I mentioned before, for me, the whole purpose and beauty of a watch is to wear it (I don't believe in a beautiful handcrafted but unwearable watch...I've seen some of those at Basel last year).
Now, about my own cal.04.0...when I received the watch, it ran about +4 sec/day, but the difference increased suddenly to +20 sec/day, after I did not wear the watch for almost 2 months...this margin of error remained for a while but again, after not wearing the watch for a couple of days, the difference reached +26 sec/day...this is a little bit strange but not very annoying.
So, I'm looking forward to read your experience regarding this matter.