I would have been more shocked if he had found no evidence of binning...
Basically this. I would be doubting the competence of any company that didn't do this, you should always send your best product out for review. I doubt Intel are testing each CPU individually for golden ones to send to reviewers as it would be just too time consuming. They are however, most likely identifying a good clocking batch of CPUs during QA testing and setting them aside for reviewers. That would cover why not every CPU is a golden sample, but most are above average, and there are a small number of poor clockers as there is still some variance within batches. I would like to know what the batch number is for the review sample CPUs to see if this is the case.
If Adored wanted something to look in to extend this topic maybe look at if CPUs from batches produced late on in in the CPUs retail cycle still, on average, overclock as well as they did when the first released. As just from my casual observations one these forums I have noticed that sometimes these later batches of CPUs don't seem to clock as well as their predecessors did.