Shoot, you're right. I miscounted the number of zeroes in the first step when copying from a python shell. All later numbers were computed from the initial values, which is why the order of magnitude was correct for the later values.