Going out on a limb here, but sfgate.com might have some bias. It's easy for data to be measured incorrectly when there is an incentive to getting it wrong.