trials = 5, don't we lose info?

Hi,

runArgs = {
"iterations" : 2000, # NULL defaults to (max available - 1)
"trials" : 5, # 5 recommended for the dummy dataset
}

My understanding on trials is that, while performing multiple trials, each trial takes different variables (not all the variables, it considers few out of it) & samples. and it picks up the best trial of it. Don't we lose information because of taking few variables (not the complete list of input variables)?
whatever variables we use in MMM all are important right. Please explain if my understanding is correct.