We can also arrive at this by looking directly at the matrix equation that defines the flexibility matrix, rather than only restating the coefficient definition in words.
The flexibility matrix relates the vector of coordinate displacements \(\{\delta\}\) to the vector of coordinate forces \(\{F\}\) as \(\{\delta\} = [f]\{F\}\). If we choose the force vector to be a unit vector with a 1 in position j and 0 everywhere else, i.e. apply a unit force only at coordinate j and nothing elsewhere, then multiplying out \([f]\{F\}\) simply picks out the \(j^{\text{th}}\) column of \([f]\), and the resulting displacement vector \(\{\delta\}\) gives exactly that column's entries.
Substituting a unit force vector directly into the defining matrix equation confirms which operation generates a column of the flexibility matrix.
Therefore, the correct answer is a unit force is applied at coordinate j and the displacements are calculated at all coordinates.