Unfortunately, at this early stage (only been available for roughly a year and only open to anyone to take for about four months) searching for matches that will help you expand your tree or get around brick walls is mostly a trial and error process.
I personally pay little attention to estimate of confidence as i have found best discoveries in low or very low confidence groups. I now search every match thru every branch, regardless of whether ancestry.com has found a matching surname. Most important is they only match 10 generations of names so they miss most of colonial american names that tend ( for me at least ) to be 12 or 13 gens back. Different for male names that passe down more generations, but those female daughters of pilgrims never show up.
On spelling of surnames, i haven't figures that out yet..i look for locations, dates, and surnames and sometimes overcome misspellings that way.
The bottom line, however, is having matches to other researchers with robust trees. Ancestry.com can tell you the dna matches, but if the other person hasn't filled out their tree, there are not names to match. It is very timeconsuming, but i search every branch of every tree i match with, even when ancestry finds a shared ancestor because, especially with colonial america, there are sometimes multiple matches.
Maybe someone will write a "dummies guide to ancestry.dna" but right now, it is pretty much looking under every branch of a tree to find those little jewels...you may find a only a handful of great links, but when it is a break thru a brick wall (often those female lines) it is super exciting.