I am reading an article and the authors at a certain point give this formula to bound a Rademacher complexity: $$ R_n(\mathcal{C})\leq\frac{M}{C_1M^{\frac{1}{p}}+C_2M^{\frac{1}{q}}}\left(\sqrt{\frac{2\ln M}{n}} + \sqrt{\frac{1}{n}}\right). $$ They then proceed to say that by setting $C_1=1, C_2=0$ and $p=4/3$, they get a complexity $O\left(M^{1/4}/\sqrt{n}\right)$. This sounds strange to me since I would say that the complexity is $O\left(M^{1/4}\sqrt{\ln M /n}\right)$.
Am I misunderstanding something?