It should be noted that Mason-Dixon and SurveyUSA always underestimate Dems in MN in Dem wave years by about 5-10 points.
So, Walz wins by about 10-15 in the end and Smith by at least 10-15 as well.
I was curious about this so I went to check.
2016: Both overestimated Dems
2014: Both were spot on
2012: SUSA overestimated Dems, Mason Dixon overestimated Reps
2010: SUSA was spot on, Mason Dixon didn't have a poll
2008: Both overestimated Reps
2006: Both were pretty close to the mark
Certainly doesn't seem to be any consistent trend there.