The answer is yes. 
Indeed, let us write $X$ instead of $x$, according to standard notation, to distinguish between random variables (denoted by upper-case letters) and their values (denoted by lower-case letters). Let us write $P$ instead of $\Pr$, and then let us also write $A$ instead of $E$, to distinguish it from the expectation sign. 
 
Then we need to show that 
\begin{equation}
	EX\,P(A|X)\ge P(A)\,EX
\end{equation}
given that 
\begin{equation}
	P(A|X>y)\ge P(A)
\end{equation}
for all $y$ such that $P(X>y)\ne0$. 

We have 
\begin{multline*}
	EX\,P(A|X)=EX\,E(1_A|X)=EX\,1_A \\ 
	=E\int_0^\infty dy\,1_{X>y,A}=\int_0^\infty dy\,E1_{X>y,A} \\ 
	=\int_0^\infty dy\,P(X>y,A)	=\int_0^\infty dy\,P(A|X>y)P(X>y) \\ 
	\ge\int_0^\infty dy\,P(A)P(X>y)=P(A)\int_0^\infty dy\,P(X>y)=P(A)EX, 
\end{multline*}
as claimed.