I've just downloaded the OCR'd version of the Department of Justice emails, and run some quick analysis on them.
Number of emails per week, April 2004 to Date.
Then I narrowed it down to just the last year
There certainly does seem to be a bit of a gap in November, but there also is a second drop in email volume in Feb 2007 when the story first started being discussed.
Steps taken to get these graphs below
There are 1032 unique "Sent: " timestamps in the data. I chose Unique timestamps to make sure that quoted text would not contribute multiple times. This then results in 186 unique dates. I then combined them into weeks and graphed the number of occurances.
Because the source for this data was a OCR'd version of the original data source, there were multiple computer generated errors (March OS, and so on) that had to be corrected by hand. This also means I wouldn't trust this data without someone redoing it from the first-generation data - i.e. the original documents downloaded from the DOJ website themselves.
Also, the fact that the OCR'd data doesn't make any distinction between the Original Sent: tag and quoted text (not the fault of jukeboxgrad, it's a limitation of the email format) there could be other hidden errors. I have taken steps to reduce the possibility of these whenever I could.