Sorry this seems to be over my head.....
You generally model a "real source" by an ideal voltage source and an internal resistor Ri. (You can also model a current source...)
When you draw current from that source the terminal voltage sadly drops down (to zero in case of a shortcut). Maximum power (=IV = II*Ri) is sucked when the terminal voltage is half of the OCV. This is a consequence of the "I square" in the formula, given that Ri is a constant.
Though this is the maximum does not necessarily mean that this is an "optimum" in a more comprehensive context.... You generally want to have current at the best possible voltage (or voltage at the best possible current).....
To add dissipating resistance (by FET or whatever) seems a foolish act at the first glance...