Suppose you have $4$ stars, then putting the stars on a diagonal is an optimal solution (however, any composition such that the rank of a $4\times 4$ matrix is full would work). That is,
$$
\left[\begin{array}{cccc}
* & & & \\
& * & & \\
& & * & \\
& & & *
\end{array}
\right]
$$
Now, if rows $R$ and columns $C$ are chosen and $R\cap C\neq \emptyset$, then there would be one star left in the matrix. For example, removing rows $1$ & $2$ and columns $2$ & $3$ would leave the star at $(4,4)$ location. However, if $R\cap C = \emptyset$, then all starts will be removed. For example, removing rows $1$ & $2$ and columns $3$ & $4$ would remove all the stars.
Suppose you remove rows $1$ & $2$, then you must have at least one star in the locations $(3,1)$, $(3,2)$, $(4,1)$, and $(4,2)$. In such a case, one can change a strategy and instead choose columns $1$ & $2$, then there must be at least one star in the location $(1,3)$, $(1,4)$, $(2,3)$, and $(2,4)$.
It can be generalized to the following: When two rows $i$ & $j$ are chosen, there should be a star in the location $(k,i)$, $(k,j)$, $(l,i)$, and $(l,j)$. When two columns $i$ & $j$ are chosen, there should be a star in the location $(i,k)$, $(i,l)$, $(j,k)$, and $(j,l)$. Since any two rows or any two columns can be chosen, we must have two extra stars to 'dodge' the row and column choices. This gives $4$ stars chosen in the beginning, $2$ additional stars, and $2$ extra 'dodgers'. So the minimum number of required stars is $8$, not $7$.
For instance, the following is a solution
$$
\left[\begin{array}{cccc}
* & & * & * \\
& * & & \\
* & & * & \\
* & & & *
\end{array}\right]
$$