Well they screwed it.

http://www.popphoto.com/article.asp?section_id=5&article_id=1881&print_page=y

"Plug in my findings for the EOS 5D and you get the following:

((2130D+2020D)/2 +2310H +2220V= 2200 average lines

That's more than enough to qualify the EOS 5D for an Excellent
resolution rating.

When we're wrong, I'll admit it
Unfortunately, 2310Hx 2220Vx 2080 Diagonal lines weren't the figures
listed for the EOS 5D in our December, 2005 test report. Those figures
were 2050V x 2000H x 2025 Diagonal, for an average resolution of 2030
lines. That's 170 lines less, or about 8% less resolution. Why the
difference? It seems that the Pop Lab technician who ran the tests on
the EOS 5D (and the Nikon D2X, also about 10% lower reported
resolution compared to recent tests) made a measurement error that
skewed the results downward. To prevent this from happening in the
future, we have added safeguards to our testing procedures.

Why bother correcting this 8% error in tested resolution for the EOS
5D, when it didn't actually lower the camera's Excellent resolution
rating? Because little errors have a way of snowballing into bigger
ones—which is exactly what happened when I used those inaccurate
resolution figures to compare the new D200 to the EOS 5D and Nikon
D2x. In that report, I noted that the D200's average resolution of
2075 lines was equal to the D2X, and slightly higher than the EOS 5D.
In fact, there's a much larger margin between the D200 and D2X (which
has been updated to 2440 lines). And at 2200 lines, the EOS 5D shows
an obvious advantage in resolution over the D200. All other test
results remain the same as reported, with the D200 showing higher
color accuracy than the EOS 5D."