Most of the time, when using MBC, the menu is considered a whole where all items are present. Thus, the simulator produces predictions for all items on the menu.
But, if you experimentally designed the MBC menu to have products that sometimes appear and sometimes disappear, then you should also have computed an "availability term" that signals to the utility estimation routine (as a dummy code) when an item is removed from the menu. When you set that "availability flag" within the simulator (just another independent variable) to indicate that the product was not available, its share should be driven to nearly (almost exactly) zero.
However, my guess is you haven't done that. My guess is you want to drop an alternative from the menu and have the remaining share for the other product alternatives sum to 100% (which is something that respondents never actually saw in the survey). If that's the case, then I don't think our Excel-based simulator that the MBC software produces can automatically do that. So, you'd need to do some modification to the Excel simulator so that the individual respondent shares of preference for the dropped product alternative are set to zero and the other items are shown to capture the full 100% of share. This makes a simplifying assumption, however, that the remaining products in the market simulator take up share in an "IIA" manner, proportional to their previous shares (it assumes constant substitution rates).