If you're using the a = (x^2+e) / (n * 2) method of generating a cell, you're diving by zero for n=0. If N is the midpoint for A and B, n=0 implies A=B. Which implies X=0, and D=A. It's outside the system of rules the VQC is inside of.
I don't think they're valid records so much as artifacts of the generation process. Notice that all the X's are the same for each N0 cell. Which means they're unindexable with our current understand of T. It might mean that D is part of the formula for T though.