Lighthouse Studio doesn't offer and option to really get in and change the coding of the tasks prior to modeling, nor do we record if a respondent chose the best on the page first or the worst on the page first to try to make the second question from a reduced set (I think this is what you are referring to for sequential marginal, where the second answer removes the first answer?).
My guess is your best bet would be to export from our software and then see if an R package can perhaps assist with changing the task codings, but perhaps someone else might have an idea.