First of all, that was over 12 years ago, so I'm not sure I remember why :-)

12 MHz allowed us to have about a 10% performance increase due to increased clock speed. The vast majority of applications (once again, this was many years ago) used 9600 baud, or even 2400 or 1200. A 12 MHz crystal works perfectly with any of these baud rates.

One big benefit was all the timing applications using the PCA where a 12 MHz crystal resulted in 1 microsecond per clock tick.