Ed,
I'm going to take a sway at this one.  Points you made:

1-Compass error worst with compass to middle or right (30 degrees or so)
while mounting to left and rear is only 10 degrees.
2-Turning on the master switch causes problem.  Pulling breaker stops
it.

From point 2, I think your problem is the loop of wire for the field
excitation of the alternator, not the alternator itself.
Also:
 -- Moving the compass to the rear seat would have worked better than it
did as the change in distance to the alternator would be much greater
than it will be to the centroid of the area of a loop of wire to the
breaker.
 --  In general the alternator should not have a substantial external
field, especially when the engine is running as the armature is the
field generator and effectively its rotation would cause any leakage
fields to be AC in nature, and thus the compass would ignore them.  (But
I am troubled by "…push the breaker back in but then higher RPM's will
cause it to swing."  This also implies that the battery cable loop is
involved as with higher RPM's less field current is necessary, but the
DC output of the alternator increases.

I would suggest the following thought experiment.  Imagine tracing the
current flow in sequence from the battery, to the master relay, circuit
breaker, regulator, alternator, frame and then by the shortest route
through the engine frame, battery ground cable and back to the battery.
Unless this entire path can be described as a coaxial cable, it will
have area enclosed by that route.  If the area is large, i.e., could you
have routed your wiring to substantially reduce that area and if the
compass sensitivity plane is substantially orthogonal to the plane of
this loop and if the compass is essentially in this plane, that would be
my guess as to the problem.  This same thought experiment should then be
done with the battery cable from the alternator to the battery and the
same restrictions apply.

You cannot shield a compass with mu-metal.  If the shielding worked,
then it would not see the earth's magnetic field.  It is potentially
possible to intercept the field from the loop effect above, by shielding
the wiring, or making it balance out.  But in practice nulling
techniques are hard to implement, and if the loop is large as I suspect
it is, then that's not very practical either.  The best technique is to
route the wiring such that the current going is balanced by current
coming back, i.e., twisted pair wiring.  Obviously not too practical
with battery cable, but the cable could be dressed close along the same
path the current will take in returning.  For instance, assuming the
regulator is mounted on the alternator, then routing sequence could be
described as: Battery, master relay, circuit breaker, back along the
line and beside the master relay, along side the battery, along side the
ground strap to the frame then along the frame out to the regulator
That will reduce the loop area, thus eliminating the extent of the
magnetic field at its source.

I hope this was clearer than mud - if not, ask questions, and I'll try
again.

Charles R. Patton
360JM