The claim seems to be that the tool will always tell you absolutely whether a null-dereference will happen, and never say “maybe.” The Halting Problem is trivially reducible to this (just put an intentional null-deref in front of each halt).
It can tell you that, "if this line is reached, and it was called from this path (with evidence on how this could happen), it will dereference a null." It is not saying that "running this program will guarantee give you a null dereference."
The halting problem is more in line with "this program will terminate on any possible inputs", which is more expansive. It is trivial to say that you can prove some programs won't terminate on a particular input. Question is if you can do it for all inputs, no?
But I don't think the Op was claiming it was possible to do this perfectly, just possible to write very useful tools. You will always have either flase positives or false negatives (or I suppose inputs where you just hang).
Turing completeness is the bane of static analysis, but that doesn't make it a fruitless endeavor.