That's not really a problem, you just a number of them you can evaluate (say, 10), either from some recommendation list or randomly, and ignore the others. The end result will be the same as if those others hadn't existed at all.