The original software is far slower than you give it credit for... I benchmarked old openssl code on a contemporary P3 a while back and got about 47KH/s as I vaguely recall. Why don't you go test it instead of making guesses from comments.

I would also not assume that the timestamps on the pre-public blocks are accurate— all we really know for sure was that there were at least two blocks created between 03/Jan/2009 and 11/Jan/2009. Rounding that down to 7 days suggests a lower bound hashrate of 14 KH/s.  Even if Satoshi had more computing power he might have simply been borrowing it for testing.

The guesswork that more work was done on the initial block is not supported by the low extranonce— you can't just say "more zeros == more work", for example:  http://blockexplorer.com/block/00000000000000001e8d6829a8a21adc5d38d0a473b144b6765798e61f98bd1d   has an "apparent" difficulty comparable to the total work ever done on the blockchain.  Sometimes low values happen.

Because its required to have at least that much, the difficulty cannot go below 1.