First, I'm guessing that's a typo in your query and the last line should read and T1.Col2 = T2.Col2; ;)
Carrie Ballinger has some good blog posts on this site about statistics and multi-column statistics. I'd recommend you take a look at those for a pretty detailed analysis of statistics recommendations.
One option is to try one approach, and check the explain plan. If you see an improvement in the confidence, great. If not, drop the stats and try the opposite approach.
If the Col1, Col2 combination is used to filter quite frequently, collect on both columns together (and look at making that an index).