\subsection{Removing a disk around each zero}\label{sec:FH} Lemma~\ref{lem:P} uses only that the image of the inverse branch of Lemma~\ref{lem:branch} at $0$, pulled back to the disk by $W^{-1}$, misses the points of $A$. In fact it misses the whole pulled-back image of the branch at each zero, and that image contains a disk of explicit radius. \begin{lemma}\label{lem:FH} For every zero $a$ of $B$, with $r=|a|$ and \[ s(r):=r_{\rm in}\,\frac{(1-r^2)(1-r)^2}{4r}, \] one has $s(r)0$, put $x_1=r-s$, $x_2=r+s$ with $s=s(r)$, and \[ p=\frac{1+x_1x_2-\sqrt{(1+x_1x_2)^2-(x_1+x_2)^2}}{x_1+x_2},\qquad q=\frac{x_2-p}{1-px_2}. \] The M\"obius map $w\mapsto(w-p)/(1-pw)$ sends $\partial\mathbb D$ to itself and the circle $|w-r|=s$ to $|z|=q$, so $\lambda_{\mathbb D_a}(0)=(1-p^2)\,\lambda_{A_q}(p)$, where $A_q=\{q<|z|<1\}$ has the density \[ \lambda_{A_q}(t)=\frac{\pi}{2t\log(1/q)\sin\bigl(\pi\log(1/t)/\log(1/q)\bigr)} . \] Since $s(r)$ is proportional to $r_{\rm in}$ and a larger disk removed gives a larger density, the bound $1/(r_{\rm in}\lambda_{\mathbb D_a}(0))$ of Lemma~\ref{lem:FH} is largest at $r_{\rm in}=1/4$; we write $\operatorname{bd}(r)$ for its value there. Every Blaschke model satisfies $K_0\le\operatorname{bd}(|a|)$ for each of its zeros $a$. \begin{corollary}\label{cor:FH} Every zero $a$ of $B$ satisfies $|a|>1/2.7051=0.369672\ldots$. \end{corollary} \begin{proof} $K_0\ge1/|a|$, since every factor $1/|a_j|$ of $K_0$ is at least $1$, and $|a|\ge\rho_*=0.31088\ldots$ by Corollary~\ref{cor:P}(b). On the interval $0.3108\le r\le1/2.7051$ the function $\operatorname{bd}$ is at most $2.7050891<2.7051\le1/r$, so no zero has its modulus there. The bound on $\operatorname{bd}$ is a ball-arithmetic computation~\cite{Johansson2017} at $200$ bits: the interval is cut into $11$ pieces, and on a piece $[r_-,r_+]$ the closed disk about the midpoint of radius $s(r_+)-(r_+-r_-)/2$ lies inside $\overline{D(r,s(r))}$ for every $r$ in the piece, since $s$ is decreasing ($(1-r^2)(1-r)^2$ decreases and $4r$ increases); a smaller disk removed gives a smaller density, so the value of $1/(r_{\rm in}\lambda)$ at that one disk bounds $\operatorname{bd}$ on the piece. \end{proof} Corollary~\ref{cor:FH} improves the bound $|a|\ge1/3$ that~\cite[Thm.~7.1]{JatarNg2026b} gives for every zero. For $\deg B=1$, Theorem~\ref{thm:B1upper} gives $|a|>0.84904$.