Does PyEphem use the complete VSOP87 planetary theory and ELP-2000/82 lunar theory or a reduced version (such as given in Meeus)?
According to the reference manual for XEphem (which pyephem is based on), "Credits" section:
The high precision planet positions were implemented for XEphem by Michael Sternberg based on the papers
- "Planetary Theories in rectangular and spherical variables: VSOP87 solution" by Bretagnon P., Francou G., in Astron. Astrophys. 202, 309 (1988), ftp://ftp.bdl.fr/pub/ephem/planets/vsop87/, and
- "Representation of planetary ephemerides by frequency analysis. Application to the five outer planets" by Chapront J., Astron. Astrophys. Suppl. Ser. 109, 181 (1995), ftp://adc.gsfc.nasa.gov/\ pub/adc/archives/journal_tables/A+AS/109/181.
See the comments in chap25.h and vsop87.h for accuracy estimates.
The high precision Moon code was also implemented for XEphem by Mr. Sternberg based on code supplied by Stephen L. Moshier at ftp://ftp.std.com/pub/astronomy/selenog.zip. Mr. Sternberg also incorporated the algorithm for deltaT, based on code also provided by Mr. Moshier. See the comments in deltat.c for full references. My greatest thanks to Messrs. Sternberg and Moshier for their generous and kind assistance in making XEphem a program of first-class accuracy.