Doug:
Your patch did the trick. Reports 2,423,793,110 points processed.
great.
That is less than reported by wc -l , but I had some upper and lower
limts to exclude outliers, which may account for the difference.
you can look at debug messages for a little more info, rerun with
scan z-bounds inverted, and check r.univar's SUM for a "method=n" map.
g.gisenv set="DEBUG=1"
you'll get more info with level 2, back to none with 0. The max is 5 but
r.in.xyz doesn't use that.
Did you want a run without limits to see if the numbers exactly match?
It's up to you.
see also http://trac.osgeo.org/grass/ticket/123
points exactly matching the southern boundary are ignored.
and the BUGS section at end of the help page.
It would be comforting to have the books exactly balance.
Hamish