This is going a little off-topic, but anyway, I made DE405 ephemeris implementation about 8 years ago and I checked it against JPL-Horizon web service and they are exact match. So, maybe we could place an other moon up there and see how much they deviate. One problem is that DE405 uses ICRF-J2000 reference frame which is a little off from the EME-J2000. Unfortunately the conversion between ICRF and EME goes over my head. I guess it wouldn't hurt to have oapi call for the conversion. Also, I need to find out if the DE405 data files can be redistributed.
Would this be a valid approach ?
Yes, this might provide a good validation. The VSOP87 and ELP82 solutions are a bit dated now (and in some cases I am even truncating the series early in the interest of performance), so it would be interesting to see how it stands up to a more modern solution.
One of the advantages of VSOP is its speed, which is important in a time-critical application where potentially a large number of bodies need to be processed. So ultimately we might need to find the best tradeoff between accuracy and performance.
Do you think you could write an alternative Moon.dll using the DE405 solution for comparison?