Inequality regarding the smallest real part of eigenvales

Define $Re\lambda_{min}(A)$ to be the minimum of the real parts of the eigenvalues of a matrix $A$. Let $A,B\in \mathbb{R}^{n \times n}$ be two matrice such that $Re\lambda_{min}(A)>0$, $Re\lambda_{min}(B)>0$ and $Re\lambda_{min}(A-B)\geq 0$. Is $Re\lambda_{min}(A)\geq Re\lambda_{min}(B)$ right? And how to prove.

Any help will be appreciated!

-
Typo in the title; typo in your final inequality –  Yemon Choi May 10 '12 at 8:14
Can you do this for $2\times 2$ matrices? –  Yemon Choi May 10 '12 at 8:14
It seems wrong: take just A=diag(10+epsilon, 1) B=diag(10,2) –  Alexander Chervov May 10 '12 at 8:19
Alexander, A-B=diag(epsilon,-1) in your example so it doesn't work. –  Felix Goldberg May 10 '12 at 8:48
@Felix, yeh, you right, I misread as abs(Re(Lambda_min))... –  Alexander Chervov May 10 '12 at 9:16

Here is a $2 \times 2$ counterexample:
$A=\begin{bmatrix}10 & 19 \\\\ 8 & 16\end{bmatrix}$
$B=\begin{bmatrix}9 & 2 \\\\ 17 & 7\end{bmatrix}$