Great question!
There isn't a formula to know for sure when to move from the setup phase to the actual questions, but you will get a sense for it after you've practiced a number of games.
I generally make the jump when I've stopped making new inferences, have spent a few moments looking things over, and nothing else jumps out at me. You will also see as you get through more game types that some games lend themselves to very limited setups which are worth sketching out up front (e.g. limited solution games, where there are a large number of rules and few uncertainties) and some games that are so open-ended up front that all you can do is write out the rules (in these game types, each question will provide an additional rule which narrows the options substantially).
There's a fairly in-depth series of articles about just this question here on the Powerscore blog:
http://blog.powerscore.com/lsat/when-to ... ogic-games
In any case, it's ok to jump into the game's questions without having made all the inferences. If something is seriously amiss in your setup, you will realize it during the first question of the set. Some inferences you will only find as you get to the end of the questions for the game.
Good luck studying!