Actually, I lied about 50 meters. For one particular point I have the difference of 121 m, and I have verified by other means that valdemar.kms.dk/trf gives the correct coordinates for that point. Your idea is pretty cool actually, I'll try that.

Thanks Kirk, but that's actually the same parameters (with slightly less precision) that what geodesy.spacecenter.dk/~geod/REF/3drefdk.htm gives, and that's quite a bit off in comparison to KMS conversion. :-(