90% confidence means anything within that range would validate their model. And even if their model was perfect, 10% chance the real value would fall outside that range. So as long as the real number (which we will will have in 2 weeks once the rolling average and final tallies are in) then, well, they're still right.