Hi Bernd,

The errors in the edge values ought not to be as serious as _that_
for the lunar occultations, except possibly for very grazing results.
And in those cases, the only way Guide could get an accurate result
would be if it had a table of lunar elevation data. If it had this,
it could render the irregular lunar limb correctly, and you might
even see stars flash in and out of visibility as they passed behind
the moon. (And again, one reason people time lunar occultations
carefully is to get better data on the shape of the moon's limb.)

Fifteen to thirty seconds corresponds to about 7 to 15 arcseconds
in the sky, or a positional error in latitude/longitude of about
7 to 15 kilometers. The most likely case would be a grazing occultation;
I dunno when you saw this event or from where you observed it, but
perhaps that's the explanation here?

-- Bill