When measuring high value resistors like these you need to be aware of the accuracy limitations of the 3456A on the higher resistance ranges. If you are trying to verify that your resistors are within 1% then it is really beyond the 3456A's resistance measurement capability.
The manual states the accuracy on the 100Mohm range as 1.8% + 1 count, which is close to the error that you are seeing. The accuracy on the 1Gohm range degrades rapidly to 16% + 1 count. Both these figures are the 90-day accuracy so you could expect worse performance if your 3456A hasn't been calibrated in a while.
I suggest you try a different approach by using the resistor under test as the upper leg of a resistive divider with a precision resistor of say 100Kohm in the lower leg. Apply a stable voltage say across the top of the divider and measure the output/input voltage ratio using the ratio measurement function of the 3456A. From the measured ratio a few simple calculations should allow you to derive the unknown resistor value.?
