This is one the thing that I found curious as well.
I worked at a company who used WebTrends. Some of the their customers installed Google Analytics alongside Webtrends and then would complain that the numbers didn't match up between the two. Sometimes it wasn't close (+/- 10%) other times it was closer, but there was always a discrepancy between the two. Of course it looked since we used WebTrends, we were inflating the numbers, when in reality, it was the exact opposite.
When I read this part of the post, I could totally see them tweaking the GA numbers so they were lower.
The reason for this would be that GA filters out bot traffic differently than other vendors. That includes legitimate bots like search engine crawlers. Web analytics are just not that accurate. The number quoted by the Xoogler in an analytics manual I recently went though is an expectation of ~10%+ wiggle.
GA is always going to show 'lower' numbers than straight logs in particular because of this.
The paragraph on G analytics is also very interesting, it never seems to line up with other tracking tools like Piwik.