Because F = ma, and m is very small at the third cutoff.
At the first cutoff, mass is the full stack minus the first stage's burned fuel.
By the time of the third cutoff, the mass is enormously less: only the third stage, with only little of its fuel left.
A rough calculation says that each cutoff has about the same acceleration:
Numbers:
Stage 1 thrust 4 * 1000 + 1000 kN.
Stage 2 thrust 1000 kN.
Stage 3 thrust 300 kN.
Stage 1 mass 180 t, including 160 t fuel.
Stage 2 mass 105 t, including 95 t fuel.
Stage 3 mass 25 + 7 t, including 22 t fuel. (7 t is the Soyuz.)
At the first cutoff,
a = F/m = (5000 - 1000) kN / (20 + 105 + 25 + 7) t = 25 m/s^2.
(The -1000 kN is because stage 2 ignited at launch, and keeps firing through the first stage cutoff.)
At the second cutoff,
a = F/m = 1000 / (10 + 25 + 7) = 24.
At the third cutoff,
a = F/m = 300 / (3 + 7) = 30.
Wikipedia's numbers aren't entirely trustworthy: for example, stage 2's dry mass plus fuel is 4 t off from its gross mass. Also, this calculation assumes that all cutoffs go instantly from full throttle to zero. So something fishy's going on. But it still seems plausible that something massing 10 t can be jostled more easily than something massing 157 t.