I am Research Engineer working in Video Analytics area and I often deal with these kind of situations. Isn't it good enough to use the existing camera network rather than upgrading to Mega-pixel cameras for using it for Video Analytics etc? If you are just trying to record video continuously without any video analytics, well I don't know what to say (old school, maybe??)
But given your experience in this area, when you are investing so much money on video surveillance, isn't it a smart idea to think ahead and implement a system that can scale well for future applications. In which case Megapixel cameras is your ANSWER!!!
When designing a video analytics algorithm (say License Plate Recognition), there are certain pre-requisites like the pixel height of the License Plate text in the image, quality of the image etc which have to be taken into consideration for accurate recognition. With 1.3MP analog cameras, it might work best if you can focus the cameras for individual lanes. Just because a human can guess the text license plate from a 1.3MP camera covering 3 lanes (which I highly doubt), doesn't mean that the computer can recognize it too. Humans have a higher level of visual perception than computer algorithms. It is a Holy Grail for any algorithm designer to replicate human vision on a computer and is still many years away.
If you can upload a sample image from the 1.3 MP analog camera covering 3 lanes of traffic and still read the text on License Plate, your arguments can be taken seriously. But I honestly don't believe that's possible.
