I like the averaging idea for user process time, but would suggest that the pictured formulation is somewhat redundant between CPU Cycles in Use and Average User Process Time. How about making Average User Process time a graphical function based directly on CPU Cycles in Use? The idea is that when the CPU is not busy, the user process time is "normal", but as the CPU gets busy (CPU cycles approach maximum available) the average user process time increases at a faster rate, indicating a "bogging down" process.
Average user process time could be a function of more things such as process mix, availability of memory, etc... This model could become quite dynamic!
Paulo is right on that the equations are correct and that the units are consistent.