process.m - difference in filter output when changing numBlocks

I've been re-creating the example for process.m from the Matlab Scripting API manual and have come across some results I'm trying to figure out. What I do is run the example in two separate ways. The first method follows exactly as the manual describes and the call to process() uses a value for NUMBLOCKS that process the entire signal at once. In the second method, I instead call process() iteratively with NUMBLOCKS=1. When I take a difference between the outputs I get some results that I can't explain:

What I'm trying to figure out is why there is a difference at all. Is the state of the filter not stored between calls to process()? Now that I think about it, I might need to clear/reset the system in between runs to make sure the trials are consistent and the internal buffers are cleared. Any input would be greatly appreciated. I will eventually take the iterative approach for a much more complex system so I should understand the fundamentals before I move on to something of a much larger magnitude.

Thanks