An arc second is one sixtieth of a nm measured at the equator. This is therefore about 31m between each of the terrain samples in the grid. So he is saying that the terrain samples are about 185m apart.
The next part makes no sense to me: "We can allow up to a 2 degree
shift and still get it right." Two degree shift in what ? I speculate that he could be saying that your that true position can be 2 arc seconds wrong, and still relate to the same terrain point.
Which function is the author describing ? It is also not clear from the information provided the sensitivity of the different EGPWS functions to the accuracy of the aircraft position.
hugel