The latest US Census Tiger data is a lot more detailed than the older Tiger data on which OSM in the USA is based. In areas in OSM that have not been updated, the roads have too few points to accurately represent their shapes. The Tiger data may have 30 points where OSM currently only has two. One could argue that Tiger actually has too many points (I have been culling a lot of them for my own project). The 2008 Tiger data also suffers from data corruption in many counties but it seems that the 2007 data does not. I am currently downloading the 2007 Tiger data to see how it looks. In addition to greater precision, the Tiger data contains vastly more streams and water bodies than the OSM data in the USA. I find that the streams are great references for location and I miss them in OSM.
I think it may be worth considering creating a way to re-merge the current Tiger data into the OSM data since it has much better precision. Even if you only added the streams and water bodies from Tiger, it would greatly increase the quality of the US OSM data.