The second problem was a little harder to figure out - I looked up
the
quartz crystal (3.93216 MHz ) which the clock
depends on for accuracy.
There is a pair of capacitors used to load the crystal and the
required value was specified at 17 pico-farad.

I found useful information on crystal timing at:
http://www.maxim-ic.com/appnotes.cfm/appnote_number/555
The first thing was that a clock running fast can be adjusted by
increasing the capactor values. The second thing
was a formula the total value for capacitors hooked up in series. If
they are the same value, the total value
is half the value of the individual capacitors.

I found the load capacitors in the clock were 15 pf which calculates
out to a load value of 7.5pf versus the 17pf
specified. I substituted two 33pf ceramic capacitors for a
calculated
value of 16.5pf - much closer to the 17pf
specified. The result was the clocks (I changed both) now gain or
loose 1 second or less in a week of testing.

You may wonder why I went to this much trouble -
1. I could not find another clock that might work.
2. I had cut a hole in my dash which the clock fitted into.


Many devices replace one of those loading caps with a small trimmer
capacitor, with careful tweaking you can get it to be quite accurate, as
good as a few seconds per month. The next problem you'll have is large
temperature swings in the car, but if you're feeling geeky you can overcome
that problem by fabricating a small oven for the crystal to sit in. This can
be as simple as a power resistor for heat, a thermister for feedback and a
simple circuit to control it, with the crystal, resistor and thermister
stuck together with heatsink grease and the assembly potted in a small block
of polyeurethane foam. With that arrangement you should be able to achieve
accuracy of a few seconds a year, but that's probably overkill.