Most NDB errors come from the radiation pattern getting bent, not from the receiver being wrong.
I have a dual-needle RMI; one needle for the ADF and one for the VOR. Flying across large expanses of flat totally boring terrain (say N France) in clear weather, the ADF is spot on, within a degree or two. But any assymetry on the ground, and this assymetry can be miles away from the NDB, and the radiation pattern changes. It doesn't matter what the receiver does; it cannot detect the distortion. There is no way to detect it. The pattern is bent before it is received.
The other thing is that an ADF is meant to be accurate only on the four cardinal bearings. In between you can get a significant error. So if you are flying on say a 360 track, and there is an NDB at your 045, it could actually be at your 055 or whatever. But the system should be accurate if you are flying directly to/from it, or exactly abeam. I don't think there is any spec on the max allowed error if away from the cardinal bearings.
I have a KR87, BTW, driving a KI229.
It's common for the NDB/ADF system to be off by 30 degrees (yes 30) due to terrain assymetry, and the error will change according to how close you are. From say 30nm it might be spot on. When you are at 8nm it might be 30 degrees off. Then when you are at 3nm it might be correct again.
NDBs are OK for navigating across vast expanses of land or sea, and are accurate for that (thunderstorms excepted). They are also very accurate as locators; flying overhead. They are much cheaper than VORs to build and maintain.
The #1 State Secret of the anti-GPS lobby is that they aren't any good for most instrument approaches - the very place where you need to be in the right place....
The most hilarious thing is that you have to carry an ADF for IFR in CAS, in UK airspace.