Another thing to keep in mind here regarding 100KHz measurement of a 100K Resistor, is at +-0.1% that's the equivalent of ~+-0.016pF (16fF) capacitance change.
Remember this is in the entire setup including instrument, DUT, any cables, fixtures and whatnot, and quality OEM Kelvin cables help but don't eliminate this potential error source. Cable bending, placement and the position of the clip tips have effects, also Kelvin techniques do not circumvent DUT shunt effects such as distributed capacitance between Kelvin clip tips which are not shielded wrt a ground shield. This is why we prefer for LCR measurements, especially high impedance and upper frequency ranges, using a test fixture specific for the type DUT, rather than Kelvin leads, even the fixture for leaded DUT like the TH26048A has proven more reliable (repeatable) than Kelvin leads in our experience.
So extreme care is necessary for precision repeatable measurements of high impedance devices at 100KHz and above for these lab grade LCR meters. On our Hioki IM3536 for example, there's an entire 13 page Chapter (10.6) dedicated to just accuracy. It's very complex with 5 variables in addition to the DUT measurement parameter and range selected. These variables coefficients are for Level, Measurement Speed, Test Cable Length, DC Bias, and Temperature, and each of these variables coefficients is specified for 7 different frequency ranges from DC to 8MHz.
Best,