\subsection{An upper bound in degree one}\label{sec:B1upper} For $0\sqrt{5/7}$ for the zero $a$ of every admissible pair of degree one, by a homotopy on the circle $|w|=|a|$. The same homotopy works on other circles. \begin{lemma}\label{lem:circlehomotopy} Let $0L(\rho)$ for every $w$ with $|w|=\rho$, then for every $W\in\mathcal S$ the derivative $(WB_s)'$ has a zero in $\{|w|<\rho\}$. \end{lemma} \begin{proof} For $0\le t\le1$ put $\widehat P_t(w)=(w-s)(1-sw)e^{t\ell(w)}+(1-s^2)w$. Since $wW'/W=e^{\ell}$, one has $(WB_s)'(w)=W(w)\widehat P_1(w)/(w(1-sw)^2)$, as in the proof of~\cite[Thm.~7.4]{JatarNg2026b}, and $W(w)/w$ has no zero in $\mathbb D$; so it suffices to find a zero of $\widehat P_1$ in $\{|w|<\rho\}$. Let $|w|=\rho$. Then $(w-s)(1-sw)\neq0$, so $\widehat P_t(w)=0$ would mean $e^{t\ell(w)}=-\sigma_s(w)$; then $t\ell(w)$ is a logarithm of $-\sigma_s(w)$, and $|\operatorname{Log}(-\sigma_s(w))|\le t|\ell(w)|\le L(\rho)$ by \eqref{eq:grunsky}, a contradiction. Hence no $\widehat P_t$ vanishes on the circle $|w|=\rho$, and by the argument principle the number of zeros of $\widehat P_t$ in $\{|w|<\rho\}$ does not depend on $t$. The polynomial $\widehat P_0(w)=-sw^2+2w-s$ has the zeros $r_0$ and $1/r_0$, and $r_0<\rho<1<1/r_0$; so $\widehat P_1$ has exactly one zero in $\{|w|<\rho\}$. \end{proof} The case $\rho=s$ is the one used in~\cite{JatarNg2026b}; there $\sigma_s$ has a pole on the circle, and the point $w=s$ is handled by $\widehat P_t(s)=s(1-s^2)\neq0$. \begin{theorem}\label{thm:B1upper} Let $W\in\mathcal S$ and let $B$ be a Blaschke product of degree one with $B(0)>0$ such that $(WB)'$ has no zero in $\mathbb D$. Then the zero $a$ of $B$ satisfies $|a|>0.84904$. Consequently \[ \mathcal B_1<\frac1{0.84904}=1.1778008\ldots<1.177801 . \] \end{theorem} \begin{proof} Write $a=se^{i\theta_a}$ with $0L(\rho)$ on $|w|=\rho$ holds for every $s$ in the range by the computation of Appendix~\ref{app:computations}. Lemma \ref{lem:circlehomotopy} then gives a zero of $(\widetilde WB_s)'$ in $\mathbb D$, a contradiction. Hence $|a|>0.84904$, and $1/B(0)=1/|a|<1/0.84904$. A pair of degree zero has gain $1$. Since $\mathcal B_1$ is attained~\cite[Thm.~8.9]{JatarNg2026b}, $\mathcal B_1<1/0.84904$. \end{proof} \begin{remark}\label{rem:valueset} Lemma~\ref{lem:circlehomotopy} uses only~\eqref{eq:grunsky} at each point of the circle, and not the relations between the values of $\ell$ at different points. Rotations of the Koebe function have $\ell(w)=2\operatorname{artanh}(e^{it}w)$ with $t$ real, so they attain equality in~\eqref{eq:grunsky} only where $\ell(w)=\pm L(|w|)$. Compare Popov's joint value region of $\log(W(w)/w)$ and $\log W'(w)$ on $\mathcal S$~\cite{Popov1965} (see the survey~\cite{Prokhorov2019}) and Grunsky's value region of $\log(W(w)/w)$~\cite{Grunsky1932}. \end{remark} \begin{remark}\label{rem:circlelimit} The proof of Lemma~\ref{lem:circlehomotopy} works on any Jordan curve in $\mathbb D$ that surrounds $r_0$ and avoids $s$, with the hypothesis $|\operatorname{Log}(-\sigma_s(w))|>L(|w|)$ on the curve. Numerically, the largest $s$ for which a circle satisfies this hypothesis is about $0.849043$ (near $\rho=0.8271$); on the circle $\rho=s$ the corresponding value is about $0.845899$. So Theorem~\ref{thm:B1upper} is close to the limit of this argument. An admissible pair of degree two with gain greater than $1.177801$ would show $\mathcal B_2>\mathcal B_1$, one of the questions of~\cite[Sect.~9]{JatarNg2026b}. \end{remark}