Extensive margin models use Rogerson (1988) style lotteries in order to simplify the savings problem under non-full-insurance.
However, one could still write down the model in that style but instead of assuming full insurance (that is, marginal utility of consumption equalized among all individuals in the household, no matter whether working or not), do it with partial insurance, in order to match the data better.
I however have not seen this approach so far. Did I miss something? Or what would be arguments against this approach? Besides the obvious one: If you don't have full insurance but forbid savings at the individual agent level, you are an evil person.