I'd forgotten about this little nugget which I had personally verified a few years ago:
How does one get a monthly and yearly mean? A monthly mean should be an average of all readings over the month. Is a yearly mean an average of all the months? No. But Hansen's source code does just that.
Why isn't a yearly mean the average of all the months?
A year is 365.25 days long. February is 28 days long most of the time. That means February readings have more weight than January readings.
That's a good point I hadn't thought of that, but I doubt it would change the results by any significant measure.