There is indeed a problem here, known as "not having updated the
Delta-T function recently enough."
The function I use had a lookup table to find Delta-T for years
from 1620 to 2002. This was based on the data from Meeus' _Astronomical
Algorithms_, with data for more recent years gathered from the following:
file://maia.usno.navy.mil/ser7/deltat.data
file://maia.usno.navy.mil/ser7/deltat.preds
I just visited these pages again, and see that some values have
been updated on the basis of more recent observations. Delta-T has not
increased as rapidly as originally expected. I had Delta-T=+64.8 for
2000; it turned out to be +63.8. For 2002, I had Delta-T=+67.0;
based on current data, it will be very close to +64.3. Also, the new
data let me extend the table to run past 2002 all the way to 2014.
Looking at the predictions and errors USNO gives for those dates is
instructive (I've clipped it to a two-year interval):
2002.00 64.26 +/- .02
2004.00 65.8 +/- .9
2006.00 67.6 +/- 2.
2008.00 68.9 +/- 3.
2010.00 70.2 +/- 5.
2012.00 71.7 +/- 6.
2014.00 73. +/- 8.
...that is, USNO is pretty sure it's got Delta-T for six months
from now nailed down to a fiftieth of a second, but has only a vague
idea what it will be in 2014.
I also have had to revise the quadratic formula used to compute Delta-T
for values past the end of the table (i.e., now past 2014) to reflect the
fact that Delta-T is apparently not increasing as rapidly as we expected
(the previous formula assumed it was gaining about 1.2 seconds/year, and
that now seems to be closer to .7 seconds/year). The result is consistent
with the new data, though admittedly not necessarily more accurate; as
the above error estimates show, once you go more than a decade or so
into the future, Delta-T "predictions" could better be called "guesses".
The fixes will appear when I next post an update. The new source code
is already posted (incidentally, it also includes the _Astronomical
Almanac_ adjustment for distant dates discussed long ago on this list):
http://www.projectpluto.com/source.htm#delta_t_cpp
-- Bill