There are no larger primes which are non-mundane than what you've indicated. To prove this, there are several cases to handle, with the following only considering odd primes. First, if $p \equiv 3 \pmod{4} \;\to\; p + 1 = 4k$ for some integer $k$, then $4$ and $k$ will make $p$ mundane unless $4 \ge \frac{p+1}{2} \;\to\; p \le 7$, i.e., the primes $3$ and $7$ you've already shown.
Next, for primes where $p \equiv 5 \pmod{8}$ with $p \ge 29$ (since $p = 5$ and $p = 13$ are non-mundame, as you've noted), consider any even integer from $4$ to $\frac{p-5}{4}$ inclusive, calling it $a_1$. Note that if its multiplicative inverse, call it $b_1$, is between $\frac{p+1}{2}$ and $\frac{3p-3}{4}$ inclusive, then $2b_1$ is congruent to a value between $1$ and $\frac{p-1}{2}$ (inclusive), so we can choose $a = \frac{a_1}{2}$ and $b = 2b_1$. Alternatively, if $b_1$ is an even value between $\frac{3p+1}{4}$ and $p-1$, inclusive, then since $2a_1 \lt \frac{p}{2}$, we can choose $a = 2a_1$ and $b = \frac{b_1}{2}$. Thus, the only possible way this prime may not be mundane is if all of the possible $b_1$ are odd integers between $\frac{3p+5}{4}$ and $p-2$, inclusive. However, the multiplicative inverse of $p-2$ is $\frac{p-1}{2}$ and that of $p-4$ is $\frac{p-1}{4}$, both of which are greater than $\frac{p-5}{4}$, so they need to be excluded.
Thus, there are
$$\frac{\frac{p-5}{4}-4}{2} + 1 = \frac{p-13}{8}$$
even $a_1$ integers, while there are only
$$\frac{p-6 - \frac{3p+5}{4}}{2} + 1 = \frac{p-21}{8}$$
odd $b_1$ integers available. Since this is one less than that of the $a_1$, and each multiplicative inverse is unique, this is not possible.
Similarly, for primes where $p \equiv 1 \pmod{8}$ with $p \ge 41$ (since $p = 17$ has $a = 3$ and $b = 6$ which work), consider any even integer from $4$ to $\frac{p-1}{4}$, inclusive, calling it $a_2$. As before, if any multiplicative inverse, call it $b_2$, is between $\frac{p+1}{2}$ and $\frac{3p-3}{4}$ inclusive, then $2b_2$ is congruent to a value between $1$ and $\frac{p-1}{2}$, so we can choose $a = \frac{a_2}{2}$ and $b = 2b_2$. Alternatively, if $b_2$ is an even value between $\frac{3p+5}{4}$ and $p-1$, inclusive, then since $2a_2 \lt \frac{p}{2}$, we can choose $a = 2a_2$ and $b = \frac{b_2}{2}$. Thus, the only possible way this prime may not be mundane is if all of the possible $b_2$ are odd integers between $\frac{3p+1}{4}$ and $p - 2$, inclusive. However, as before, we can exclude $p-2$. Thus, so far, we have
$$\frac{\frac{p-1}{4}-4}{2} + 1 = \frac{p-9}{8}$$
even $a_2$ integers, with up to
$$\frac{p-4 - \frac{3p+1}{4}}{2} + 1 = \frac{p-9}{8}$$
i.e., the same number of odd $b_2$ available. Since $p\equiv 1\pmod{8}$, it's of the form
$$p = 2^{c}d + 1$$
where $c$ and $d$ are positive integers, with $c \ge 3$ and $d$ being odd. If $d \ge 5$, then note that
$$\begin{equation}\begin{aligned}
p - 2^{c+2} \gt 1 \\
4p - 2^{c+2} \gt 3p + 1 \\
p - 2^{c} \gt \frac{3p + 1}{4}
\end{aligned}\end{equation}$$
Thus, $p - 2^{c}$ is one of the odd integers under consideration. However, its multiplicative inverse of $\frac{p-1}{2^{c}}$ is odd, so it's not included within the set of $b_2$ values, which means this set now has a size one less than that of the set of $a_2$ values, which is not possible. This means that these primes are all mundane.
Next, consider if $d = 3$. We then have
$$2p + 1 = 2(3(2^c) + 1) + 1 = 3(2^{c+1} + 1)$$
If $2^{c+1} + 1$ is non-prime, since $c \ge 3$, then $2^{c+1} + 1 \gt 3^2$. Thus, let $a$ be $3$ times the smallest prime factor of $2^{c+1} + 1$, with $b = \frac{2p+1}{a}$, with both $a$ and $b$ being less than $\frac{p}{2}$. Otherwise, $2^{c+1} + 1$ is a Fermat prime, so $c + 1$ is a power of $2$, i.e., there's a positive integers $k \ge 2$ and $m \ge 1$ where
$$c + 1 = 2^k \;\to\; c = 2^k - 1 \;\to\; c = 4m + 3$$
However, we then have
$$\begin{equation}\begin{aligned}
p & \equiv 3(2^{4})^{m}(2^3) + 1 \pmod{5} \\
& \equiv 3(1^{m})(3) + 1 \pmod{5} \\
& \equiv 0 \pmod{5}
\end{aligned}\end{equation}$$
but $p \gt 5$, so it can't be a prime.
The final case to handle is $d = 1$. This then means that $p$ is a Fermat prime, i.e., $c = 2^{k}$ for some integer $k \ge 2$. We then have
$$\begin{equation}\begin{aligned}
2p + 1 & \equiv 2(2^{2^k} + 1) + 1 \pmod{5} \\
& \equiv 2(1 + 1) + 1 \pmod{5} \\
& \equiv 0 \pmod{5}
\end{aligned}\end{equation}$$
Thus, we can choose $a = 5$ and $b = \frac{2p+1}{5}$, with both being less than $\frac{p}{2}$, to show these primes are also mundane.