As Daniel Fischer said: if this is how positivity is defined, and if the ground field is $\mathbb R$, then the statement is false. Just let $B'$ be the zero operator, and let $B$ be any non-self-adjoint positive operator such as $$\begin{pmatrix} 2 & 1 \\\ 0 & 2 \end{pmatrix}$$ The authors either implicitly switched to the more common definition of positivity (which requires self-adjointness), or omitted a hypothesis.