The TEG sees as its load the voltage across Cin.
If Cin is large enough to reduce the switching effect of the boost converter to a small ripple, then there is simply a steady(ish) current flowing through Rteg to the steady(ish) voltage on Cin. That is resistive.
The boost converter should then contrive to draw (on average) this current out of this voltage on Cin.
The problem is that this circuit does not have access to Vteg, so does not know what Vteg/2 is. There are a number of solutions -
a) hunt around for the voltage on Cin that gives max power output (this is the conventional closed loop MPPT algorithm)
b) assume conditions will change slowly at the TEG, and switch off the boost converter briefly from time to time to let VCin rise to Vteg, to measure it. This loses little power throughput if the switching frequency of the boost converter is >> the Rteg.Cin time constant, to minimise the fraction of time the boost converter spends off
c) assume a value for Rteg, and control the mean current the boost converter takes from Cin to be equal to Vin/Rteg. This will work for both DCM and CCM.
d) you might want to reference the paper from your other post in your question, as maybe controlling the ON time was a clever way to do (c)? I still haven't figured out why they think it should work, but somebody else may have more time to spare.