I guess the idea behind this is that the optimal team is the one which has, at the start, the most "hidden value"/ upside potential - that means all the rookies and midpricers that will have a massive jump in their average output during the season.
I you "force" your initial team to be worth 10M (or close - as it seems to be), you probably force the selection of premiums which don't have this much upside potential.
If the starting team price is not a constraint, it would be interesting to see if, round after round, the rolling perfect team starting value converges toward the 6/7 M total as mentioned by CouchPotatoe....