In estimating the cost per unit using a regression analysis, it means that depending on the number of units produced or the number of services performed the cost remains constant, even though this is not necessarily the case. Think about the even cost of electricity, it fluctuates depending on the time of day, time of year, how much usage, and whatever the electric companies feel like charging.
In the example in the book when plotting the Units to a calculated average cost per unit, you can clearly see that the more units produced the lower the cost per unit.
When forecasting future costs I would use the linear equation provided by excel when I found the line of best fit in the above scatter graph. Which is y=-.1061x+307.51 where X= the number of units.