Retrosheet Question

I'm looking at the Retrosheet data for 2007 and it seems like they've missed a lot of pitches.   I'm seeing a lot of places where the number of pitches in the pitch sequence is less than the ball/strike count. 

This seems to be a consistent problem.  I count over 20,000 PAs every season I've checked (1997-1998, 2004-2007), or roughly 10%, where the number of pitchers in the pitch sequence is fewer than the ball/strike count.

Even back in 1984, I get 2600 PAs, but that's not as big an issue (1.5%).

Anybody else run into this? I'm worried it makes it impossible to trust any analysis using the pitch sequence.