Why does Ashcroft and Mermin say an acceptor level in a semiconductor cannot be empty, only singly or doubly occupied? On page 581 of Ashcroft and Mermin, Solid State Physics:

Acceptor Level In contrast to a donor level, an acceptor level, when viewed as an
electronic level, can be singly or doubly occupied, but not empty. This is easily seen
from the hole point of view. An acceptor impurity can be regarded as a fixed, negatively
charged attractive center superimposed on an unaltered host atom. This additional
charge $-e$ can weakly bind one hole (corresponding to one electron being in the acceptor level). The binding energy of the hole is $\epsilon_a - \epsilon_v$, and when the hole is
"ionized" an additional electron moves into the acceptor level. However, the con­figuration in which no electrons are in the acceptor level corresponds to two holes
being localized in the presence of the acceptor impurity, which has a very high
energy due to the mutual Coulomb repulsion of the holes.
Bearing this in mind, we can calculate the mean number of electrons at an acceptor
level from (28.30) by noting that the state with no electrons is now prohibited, while
the two-electron state has an energy that is $\epsilon_a$ higher than the two one-electron states.
Therefore
$$
\langle n\rangle
=
\frac{
2e^{\beta\mu} + 2e^{-\beta(\epsilon_a-2\mu)}
}{
2e^{\beta\mu} + e^{-\beta(\epsilon_a-2\mu)}
}
\,.
$$

In the quote, $\epsilon_a$ is the energy of the acceptor level, $\epsilon_v$ is the energy of the top of the valence band, and $\langle n\rangle$ is the average occupancy of the acceptor level. Equation (28.30) is
$$
\langle n\rangle
=
\frac{
\sum N_je^{-\beta(E_j- \mu N_j)}
}{
\sum e^{-\beta(E_j- \mu N_j)}
}
$$
where $E_j$ and $N_j$ are the energy and number of electrons in state $j$.
It seems like the two electron state corresponds to a situation where the acceptor ion has a full outer shell (hole filled), whereas the one electron state is when the hole is not filled. I'm not sure what the zero electron state refers to. Why can we not say then that the situation where the acceptor hole is filled is the single electron state?1
Also, if the acceptor level cannot be empty, at absolute zero is the Fermi level between $\epsilon_v$ and $\epsilon_a$? If so, would this not mean the acceptor level is empty?

1. Like the description on page 580 "The hole is bound when the level is empty."
 A: One should remember that each energy level has two spin states. You are right in that "two electron state corresponds to a situation where the acceptor ion has a full outer shell (hole filled), whereas the one electron state is when the hole is not filled". That one electron is assumed to be occupying the corresponding bound energy state. Every such energy state can accommodate two electrons with different spins.
Having two holes attached to the acceptor impurity is energetically unfavorable, similarly to how it is unfavorable to  have two electrons on the same donor site: the negative ion is shielded by the first hole (which is assumed to be already there) and the second hole is not going to be attracted to the impurity strongly enough to form a bound state (this is a bit too handwavy argument to my taste, but I am quite sure this is what Ashcroft and Mermin mean).
The equation for the mean number of particles $\langle n\rangle$ is obtained from the assumption that there are two one-particle states and a single two-particle state.
