I look at the lower scoring models in a situation like this only as validation to the top scorer model predictions. If they tend to agree then the forecast it is that much more solid. If they do not, it's just a net neutral to the forecast. They are lower scorers for a reason.