If the tree rings were all the same width then there is nothing against which to measure any correlation.
The more measurable differences exist between the ring widths, the more possible a correlation becomes.
The less any pattern repeats the better, etc. Then the tree ring patterns become analogous to fingerprints.
So differences between individual rings from the average is not the issue except that the differences exist in the first place. The issue is the pattern of differences across the specimen and some way to characterize its likelihood of being duplicated for a different time period.
It would seem that the amount of fluctuation and variation allowed would be known from studying rings from living coexistent trees from the same area. Comparing this variation to mismatched samples would yield a control for purposes of correlation.
Now the correlation with C14 signatures is significant since if the trees came from the same area and the rings are indeed from the same time snapshot they will have the same C14 signature however C14 might vary with time. The only way correlation with C14 would be insignificant would be if there was no variation of C14 with age at all, analogous to tree rings with no variation of width at all.
So now for two ring patterns to accidentally repeat as well as the C14 signature to accidentally repeat becomes highly improbable. When the correlation of both of these patterns with varve layers is added we see that the correlation of agreement is what destroys the YEC viewpoint.