ms=3.54019 g
Which is, of course, correct. Now in this case using the Henderson-Hasselbalch would have worked fine (to my calculation, it results in an answer of 3.54068 g), but that's not as accurate as it could be.
3.54068 g vs 3.54019 g. You have just proven it is almost perfect. If you forgot - whole discussion started with your claim:
If you're using the Henderson-Hasselbalch approximation then activity coefficients are unlikely to be relevant - using the approximation will already take you 0.1-0.4 pH points away from a true value.
Yes, HH equations has its limitations, but using more complicated method when it is not needed is rarely a good thing. HH equation works pretty nicely for reasonably concentrated solutions and pH somewhere between 3-11 - and these are parameters of most of the buffers used in the lab practice.
I was forced to use that method because I didn't have any problems designed to be solved by the Charlot equation and not by HH at hand. As I clarified twice in that last post, this particular sum was likely to yield decent results from HH because it was from a high-school textbook. I already made this very clear! The HH still makes two large approximations that the Charlot does not:
Ca and Cs >> [H+] >> [OH-]
Which simply does not need to be true. Try a very low concentration acid and the HH will flop. You say it's "a more complicated" method, but again like I said it requires you to plug in just 6 extra numbers!