JS Divergence Equation Clarification

GAN-VS™

When comparing the real data and GAN-generated samples, was there a reason Jensen-Shannon (JS) divergence was used instead of Kullback-Leibler (KL) divergence? For context, I'm referring to page 81 in "Empowering DTCO Innovation with AI and Machine Learning." I see that the JS divergence is defined as D_JS(P||Q) = 0.5 D_KL(P||M) + 0.5 D_KL(Q||M). But would it be possible to just use KL divergence instead of JS divergence?