Answer: Yes, here is the proof that an = cos nq,
where q = cos-1 1/3.
Proof, by induction:
True for n = 0 and n = 1, because cos 0=1 and cos q=1/3.
2/3 cos n q - cos (n-1) q
= 2/3 cos nq - (cos nq
cos q + sin nq sin q),
from the sum of cosine identity
= (2/3 - cos q) cos nq
- sin nq sin q
= 1/3 cos nq - sin nq sin q,
because cos q = 1/3
= cos nq cos q - sin
nq sin q,
again because cos q = 1/3
= cos (nq + q),
from the sum of cosine identity
= cos (n+1)q,
so by induction an = cos nq for all
n.
If q/p
is rational, an is periodic, so there exists an n such that an=a0=1;
And if q/p
is irrational, nq is dense in the circle so an
is dense in [-1,1].

This proof doesn't depend on q/p
being irrational. It's not hard to show that q/p
is irrational, though.
2 cos nq is a monic polynomial in 2 cos q
with integer coefficients (obvious if you think about it, since 2 cos q
= eiq + e-iq).
Hence if 2 cos nq is 2 for any n > 0, 2 cos q
has to satisfy a nontrivial monic polynomial equation over
,
i.e. 2 cos q is an algebraic integer. 2/3 is
not an algebraic integer (it's not an integer). So an is never
equal to 1, and thus q/p
is irrational. (Corollary: The only rational multiples of p
that have rational cosines are integer multiples of p/3.)

As for the interesting patterns found in the investigation,
cos-1 1/3 = q = 1.23095941734077
51q
= 62.7789302843795, and 10(2p) = 62.8318530717959,
so 51q is just 0.08% less than 10 trips around
the circle.
46q = 56.6241331976756, and 9(2p)
= 56.5486677646163, so 46q is just 0.13% more
than 9 trips around the circle.
By interspersing 54 groups of 46q,
(about 9 trips around the circle) with 77 groups of 51q,
(about 10 trips around the circle), we get a total of 6411q,
which is almost exactly (to within 0.000001%, or one part in 100 million)
1256 trips around the circle.
The ratio of the number of q's
to the number of trips around the circle, 6411/1256=5.1043, is the primary
observed period of this sequence.
The weighted average of 54 groups of
46q and 77 groups of 51q
(131 groups in all) needed to make all these trips around the circle is
about 6411/131=48.9389, which is the second-order period that was observed
in the investigation.