Matasović seems reliable, but I see your point. I'll refrain from adding descendantless reconstructions in the future.