Well, good bracket making is about picking smart upsets. I took Xavier > Duke, not only because Xavier had a moderate chance of beating Duke, but *also* because WVU had a good chance of beating Duke, THEN Xavier would take out WVU. Easier said than done, eh? But dead-end upsets are pretty pointless to do, and a bracket isn't about being perfect its about probability of a given team winning in a certain slot, against any matchup it may be. That is why I often work backwards, instead of "well if this happens, then this could happen, then this could happen, so omg 13 seed in the final 4!!!"
If you agree with 99%, 90%, 80%, 70% for odds of the #1 seeding winning their round 1-4 games, on average, then every 16 years you'd see the #1 seeds in the final 4. I think the main problem is poor seeding often makes #1 seeds close to, even blow, tossups in round 3 and 4, and plenty of teams are underseeded in the 4/5 or 8/9 even, to really screw with those.