Let me have another go...
First, note that stopping power is really ##-\frac {dE}{dx}## but the authors have dodged the minus sign by referring to stopping power as the ‘rate of energy loss’. I’ll stick with their usage.
The (empirical) relationship between stopping power and velocity might best be deduced by considering Fig. 3 and Fig. 4. together. Remember Fig. 4 is directly derived from Fig. 3, being essentially a graph of Fig. 3’s gradient vs. absorber thickness. Note ##\frac {ΔE}{Δx}## has been used in Fig. 4 rather than ##\frac {dE}{dx}##.
The kink in Fig. 3 and the peak shown in Fig. 4 are anomalous, as explained at the end of the paper’s 2nd paragraph. So consider absorber thickness (x) up to x=36cm only; i.e. exclude the kink and peak
Consider thickness x=3cm.
From Fig. 3, E ≈ 5.2 MeV so velocity ≈ ##\sqrt {5.2} = 2.3## (funny units).
From Fig. 4, stopping power = ##\frac {ΔE}{Δx}## ≈ 1.0MeV/cm.
Consider thickness x=36cm.
From Fig. 3, E ≈ 1.3MeV so velocity ≈ ##\sqrt {1.3} ≈ 1.1## (funny units).
From Fig. 4, stopping power = ##\frac {ΔE}{Δx}## = 2.0MeV/cm.
Observe that up to x=36cm, the Fig. 3 graph is monotonically decreasing and the Fig.4 graph is monotonically increasing. So we see the general rule: stopping power increases as velocity decreases.