You'll see the test produce the exact same result except for these two:
AWL=0.356, BAYES_00=-2.599
vs
AWL=-0.440, BAYES_05=-1.11
The first is the auto-white list -
http://wiki.apache.org/spamassassin/AutoWhitelist
You can read from the wiki that it will vary over time and try to push the score in the direction of it's most recent scores. So that guy is just trying adjust for the difference in the BAYES_X scores. The Bayes tests will vary by the headers as Bayes tokenizes everything. You can read more here:
http://wiki.apache.org/spamassassin/BayesFaq
The bottom line is it's no big deal since both messages are clearly good mail and they result is such. What your seeing is just a little fluctuation in SA and this is expected and why there are so many tests ran on each message before something is declared as spam or not.