OK - then I can skip ther hand-wavey "pushing current" ideas ... just do a mesh analysis on the circuit.
Using the left hand diagram ...
there are two nodes - n1 is at the R1/R2 intersection and the other is at the bottom of R2.
there are two meshes ... left and right.
Let the current through R1 into n1 by I1, and the current out of n1 through R2 be I2 ... the current into n1 from the galvanometer is I3.
Then for n1: I1+I3=I2
For the left loop, going clockwise, Vs=I1.R1+I2.R2
For the right loop, going clockwise, Vr=I2R2 (galvanometer has no resistance)
Since R1 and R2 come from a potentiometer with total resistance R, R1+R2=R
Moving the slider adjusts the ratio of R1/R2 - you adjust until I3=0 ... so R1/R2 should depend on Vr and you have calibrated the equipment.
From the mesh analysis you should be able to see how I3 varies with R1/R2.
How would this work out if you put Vr the other way around?
Repeat for the uknown EMF - call it U.
But you should be able to see it from inspection if you suppose that Vs > Vr and Vs > U ... if the volt-drop across R2 is the same as U, then the current through the galvanometer is zero.