I understand “hit the $50$-use mark” to mean that you can tell when the $50$-th use has used up the roll, and you don't need to pick it one more time to find that it's empty.
The probability that the other roll has $0\lt k\le 50$ uses left is
$$
\mathsf P(K=k)=2^{-(100-k)}\binom{99-k}{49},
$$
since in that case you chose the now empty roll any $49$ times out of $99-k$ times and then once tonight.
I assume that you're implying that you haven't hit the $50$-use mark on the other roll yet, so we want the conditional probabilities $\mathsf P(K=k\mid K\gt0)$. Fortunately, we have $\mathsf P(K\gt0)$ by symmetry, so for $k\gt0$ we have
$$
\mathsf P(K=k\mid K\gt0)=\frac{\mathsf P(K=k\land K\gt0)}{\mathsf P(K\gt0)}=\frac{\mathsf P(K=k)}{\mathsf P(K\gt0)}=2\mathsf P(K=k)\;.
$$
Thus
\begin{eqnarray*}
\mathsf E[K\mid K\gt0]
&=&
\sum_{k=1}^{50}k\,\mathsf P(K=k\mid K\gt0)
\\
&=&
\sum_{k=1}^{50}k\cdot2\cdot2^{-(100-k)}\binom{99-k}{49}
\\
&=&
\frac{315285451704888104171289053925}{39614081257132168796771975168}
\\
&\approx&
7.96\;,
\end{eqnarray*}
in agreement with your simulation.
Alternatively, you could get the factor of $2$ by saying that the last use emptied either of the rolls, so you don't get a factor of $\frac12$ for the last use, and then using $2^{-(99-k)}$ for the probability of a certain pattern of choices of that or the other roll.