Why did you decide to use the same number of items each time. If you did one million simulations, each with a random number of random items, wouldn't the results have been a lot different?
There probably would be a difference, but it may be more due to the very limited sample size of items than to the random number of items selected. A real grocery story has thousands of items, but the sample list I had only had about 60-something. To get a truly accurate test it would be necessary to load a much larger list of sample items and then loop through with a randomized list of number of items selected. But the quickie test was enough to satisfy my curiousity.