First, I don't have the original document so I am going on your description only.
The secret to your dilemma lies in your observation "I got roughly the same answer of 90kg of fuel can be added".. Running the calculation graphically is fine in practice but the accuracy and precision are, necessarily, a tad rough. We address that by plotting/reading conservatively to ensure that the answer does, in fact, put you inside the envelope.
If, on the other hand, you run the sums, you will get a more precise answer than that obtained graphically .. that's the nature of arithmetic. Precise but, in practice, no more accurate as the starting empty weight values are not to that sort of accuracy in any case.
So, in effect, what you are comparing, is something akin to measuring a dimension with a school rule, on the one hand, and with a micrometer, on the other. The two approaches just won't/can't give you the same sort of answer precision albeit that the functional accuracy may well be similar, depending on the exercise details.
Checking the graphical answer via the calculation is fine and beaut but (assuming you check the working and you are comfortable that you have done the deed correctly) you should use that check to refine the value read off the graph.
We can do the graphical exercise to the same precision as the calculation check, but that involves solving simultaneous equations (one linear, one quadratic) and that is just not what the examiner wants to see. So, the best you can do is one of the following
(a) run the graphical solution and then fudge it a bit using the calculation .. all the while making sure that you finally end up inside the envelope
(b) start as (a) but then run the calculation iteratively to come up with a similar result.
Would the correct answer be C? Because if we add 90kg ( B ) of fuel its outside the limit?
If 90kg puts you outside the envelope, while 80kg puts you inside, the choice is a programmed decision ...
Rounding is ALWAYS done in the conservative direction.