Basically, all Nick is saying is that the rocks do, in fact date to 5,000 years old. However, some radiometric measurements are, by the very nature of scientific experimentation, a bit off (in much the same way, I might weight 225 lbs, but step on three scales and see weights of 225, 226, and 228 lbs). Similarly, a rock that is actually say, 5,267 years, will have radiometric measurements that do not exactly match this age.
Back to my scale analogy, it is feasible that with ENOUGH different measurements on ENOUGH different scales, I'll see some sporadic measurements that greatly differ from my actual 225 lbs (e.g. a scale with a few rusty springs reads "300 lbs"). In fact, if I repeat my experiment enough times, in theory, there is NO UPPER BOUND above which a "noisy measurement" would not be read; nor is there a limit on the number of times a reading will exceed any arbitrary point. In more concrete terms, with ENOUGH measurements, a "> 300 pound" measurement will be recorded, multiple times.
Now, enter the Blinded-by-dogma scientist Dr. JM. He, due to his philosophical dogma, has decided (before performing any experiment), that I weigh 300 pounds. So, he takes 10,000 measurements; and sure enough, the 4 highest measurements are 260, 305, 310 and 330 lbs. Though all of his data, collectively, gives a reading of 225.342 +/- .005 lbs, JM decides that 9,996 of those measurements are "noise", discards them, and concludes the experiment has weighed me in at 301.25 +/- 30 lbs.
We give JM the benefit of the doubt that filtering out the 225-lb measurements was an unconcious mistake - after all, if he'd already decided that I weighed 300 lbs, it's only natural that he would see a 226 lb measurement as, well, ludicrous. If the entire scientific community were similarly blinded by the "Baloo42 weighs 300 lbs" philosophy, it is understandable that JM's results would get published, and anyone (like a lab tech) trying to publish the "actual data" would be laughed out of the community.
Similarly, Nick is simply saying that Joe Meert, blinded by his "this rock is exactly 500 million years old" philosophical dogma, took enough measurements of a 5,000 year old rock, and eventually got some measurements of 6,000 years, a few at 7,000 years, etc, etc. With enough patience and work, Joe was finally able to compile a large number of measurments a bit larger than 7,000... even several measurements that exceeded the 5,200 year-old age by 5 orders of magnitude. (In my analogy, this would be JM eventually getting several readings of 15,000 tons - about 1/10th the weight of the Sears Tower). Statistically speaking, yeah, he'd have to perform about 10^10^100 or so experiments to get these readings, but working quickly, it's not unreasonable to guess he was able to do this over a period of time just slightly longer than the age of the known universe. And so, at the end of the proverbial day, Joe simply swept the 9,999,999,999,999,999,999,999,999,999,999,999,999,999,999,999...[,999]^10^99...,999,999,999,999,982 "bad readings" of about 5,200 years under the carpet, jotted down the 18 "good readings", concluded that the evidence supported an approximate age of 500 million years, and published his results.
Again, it's not that Nick is accusing Joe, or any other "500 million year-old-rock" dogmatic scientists of being dishonest... he's saying they just mistakenly discard the wrong set of results from a given experiment. And since the "good readings" were discarded, it is ridiculous to ask for evidence they ever existed (the scrap of readout containing the good readings would probably be as small as, say, the Milky Way galaxy - and Joe could have hidden it anywhere, like in the middle of the Orion constellation or something).
Which pretty much leaves us at an impass - two totally rational interpretations of the same data, and all lurkers able to decide which most closely fits with their worldview...
Now, can't we all just get along?