1: \begin{proof}

2: It suffices to show $\pr ( \Vert BX \Vert_1  \neq \langle X, \otherB X \rangle ) \leq n e^{-1/(5 \lambda)}$.  The idea is that since each row of $B$ is dominated by a single large entry (namely $b_{i,i}$), each entry of $BX$ is a random sum dominated by a single large term (namely $X_i b_{i,i}$).  Thus, it is very unlikely that any entry of $BX$ would have a different sign than $X_i b_{i,i}$.  This is made rigorous as follows.

3: \paragraph*{}Recall that we ordered the columns of $B$ so that the $(i,i)$-entry is the largest in its row, and that $b_{i,i} \geq 1- \lambda$.  Letting $Y_i$ be the $i^{\text{th}}$ coordinate of $B X$, we have, by a simple union bound,

4: \[

5: \pr ( \Vert BX \Vert_1  \neq \langle X, \otherB X \rangle ) \leq  \sum_{i=1} ^{b} \pr( |Y_i|  \neq X_i Y_i ) = \sum_{i=1} ^{b} \pr(X_i Y_i < 0) = \sum_{i=1} ^{b} \pr \left(\sum_{j=1} ^{n} X_i X_j b_{i,j} < 0 \right).

6: \]

7: Using the fact that for any given $i$, the random vector $(X_i X_j)_{j \neq i}$ has the same joint distribution as $(X_j)_{j \neq i}$ (and that $X_i ^2 = 1$), we obtain by \myRef{Proposition \ref{hoeffding}}

8: \[

9: \sum_{i=1} ^{b} \pr \left( \sum_{j=1} ^{n} X_i X_j b_{i,j} < 0 \right) = \sum_{i=1} ^{b} \pr \left( b_{i,i} < \sum_{j \neq i} ^{n} X_j b_{i,j} \right) \leq \sum_{i=1} ^{b} \exp \left[ \dfrac{- b_{i,i} ^2}{2 \sum_{i\neq j} b_{i,j} ^2} \right].

10: \]

11: Since $b_{i,i} \geq 1-\lambda$ and $\sum_{j} b_{i,j} ^2 \leq 1$, this in turn is bounded by

12: \[

13: \sum_{i=1} ^{b} \exp \left[ \dfrac{- b_{i,i} ^2}{2 \sum_{i\neq j} b_{i,j} ^2} \right] \leq n \exp \left[ \dfrac{- (1-\lambda) ^2}{2 (1- (1-\lambda)^2)} \right] \leq n e^{-1/(5 \lambda)},

14: \]

15: where the last inequality is justified because $0 < \lambda < 0.1$.

16: \end{proof}

17: