Very accurate, actually. The trick is not to observe the Sun directly, but to observe the other planets, and to find a self-consistent orbital solution from these observations that agrees with General Relativity. One can even test for departures from GR and for unmodeled mass.
Once one does that, one can extrapolate the planets' motions several million years both forward and backward. When one does that, one finds that the planets' orbit orientations do loop-the-loops, and that their eccentricities and inclinations oscillate quasiperiodically. This leads to Milankovitch climate cycles on the Earth and likely also on Mars.
In particular, I've plotted
Eccentricity / Runge-Lenz vector: {e*cos(w), e*sin(w)}
Inclination / north-pole vector: {sin(i)*cos(W), sin(i)*sin(W)}
for the last several million years.
I've made videos for YouTube:
http://www.youtube.com/my_playlists?p=D0825FC30A2F00A6
http://www.youtube.com/my_playlists?p=86F2CCA7F3F677ED
I got the numbers from here:
J. Laskar
"Secular evolution of the Solar System over 10 million years"
Astronomy and Astrophysics, 198, 341-362 (1988).
http://adsabs.harvard.edu/abs/1988A&A...198..341L
I had to OCR the numbers and then painstakingly correct the OCRing, so there might still be some typos.