An extension of the Weil-Carlitz-Uchiyama bound (Q1897205)

The authors claim to provide a much improved bound for exponential sums: For \(f(x)\in \mathbb{F}_{q^l}[x]\) with \(q\)-degree \(D\) and \(f\neq g^p-q+\) constant the exponential sum \[ S(f)= \sum_{x\in \mathbb{F}_{q^l}} (e^{2\pi i/p})^{Tr} \mathbb{F}_{q^l}/ \mathbb{F}_p (f(x)) \] satisfies the upper bound \[ \bigl|S(f)\bigr|\leq(D-1) q^{l-(1/2)}. \] However if we consider for \(\mathbb{F}_{q^l}\) with \(q=2^m\) and \(l=2\) the polynomial \(f(x)= x^{q+1}\), the upper bound (with \(D=2)\) yields \(|S(f) |\leq q^{3/2}\). This leads to a contradiction since \(S(f)=q^2\). The culprit seems to be Lemma 1 of the paper.