Joe's sales records for the sale of tomatoes would be the basis of establishing whether linear regression applies. His business is at least in part selling tomatoes, and there is a direct relationship between the number of tomatoes he sells and his profit. Written into this is the potential loss per month of giving away tomatoes he could have sold. So the 5 tomatoes are worth 3.75 a month. He does sell more tomatoes than he gives away. We have to assume that he doesn't lose a significant number of tomatoes as a result of decay or theft or taking them for himself or his family. We could assume that he gives away tomatoes because they would otherwise decay.
The graph of tomato sales would take the form of time against number sold or the revenue from the sales. Time could be in days or weeks, depending on when he did his accounting. He could also plot his production on the same time scale to see what proportion of the tomatoes he grows are sold. Over a period of several months it should be clear whether there was a linear trend, i.e., the data followed a roughly straight line, some data below, some data above the line in the form y=ax+b+e where a is the slope, b the y intercept and e an error factor caused by data fluctuations.