Ah, so you're just picking the two points furthest away from each other (in terms of activity) and applying the slope and point-slope formulas? That makes sense, and I assume to get a better estimate you can just do a full linear regression using all the data.
I know this is 5 months later, but usually you would go with the fixed cost that is more relevant to your problem. Most of the time a problem will not have this happen, but this could be the case in a real life situation. It all depends on your judgement to which one is more relevant to the organization.
