Yes I agree with Eric - you need to convert to decimal degrees.
The error that you are quoting is about right - I would expect errors of "upto a few tens of miles".
Once you have that fixed, if you still see an error but it is much smaller - about 100m or so, and roughly consistent across a region (eg. all locations in a city are shifted about 10-150m in the same direction), then you probably have a geoid problem.
MapPoint uses WGS84 (the same as that used internally by the GPS satellites) - but we'll tackle that problem when/if we come to it
Richard