Proposition 11: Denominator Typo

Parts (ii)–(iii): the first term is ϵγ2/(1−ϵ)2\epsilon\gamma^2/(1-\epsilon)^2 in place of ϵγ2/(1−ϵ)\epsilon\gamma^2/(1-\epsilon), where ϵ\epsilon is the recovery tolerance and γ\gamma bounds each vector norm.

For 0<ϵ<10<\epsilon<1, recovery gives |⟨bi,aj⟩|<ϵ|\langle b_i,a_j\rangle|<\epsilon for i≠ji\ne j and ∥ai∥,∥bi∥>(1−ϵ)/γ\|a_i\|,\|b_i\|>(1-\epsilon)/\gamma. Normalizing by two norms therefore gives

|⟨bi∥bi∥,aj∥aj∥⟩|<ϵγ2(1−ϵ)2. \left|\left\langle\frac{b_i}{\|b_i\|},\frac{a_j}{\|a_j\|}\right\rangle\right| <\frac{\epsilon\gamma^2}{(1-\epsilon)^2}.

The missing factor is a local normalization correction in the printed proof. It makes the finite bound slightly larger; at γ=1\gamma=1, the near-orthogonality conclusion as ϵ→0\epsilon\to0 follows with either expression. The corrected expression is proved; this normalization argument alone does not rule out a sharper bound by another proof.

Source: Proposition 11(ii)–(iii) and its proof, Section 4.