So the documentation fairly clearly states that output one is an array containing the minimum values, and the second output is the index of the first occurrence of each minimum value.
What might not be immediately obvious is that output two is exactly one dimension lower than the input (as opposed to 2 dimensions reduced). In your 2D example, "c" is a row vector that gives the row number of each column's minimum value location. You don't need a vector of column indices, because the column position in "c" is the column index of that min value. You're guaranteed a minimum value for each column, so the c array you wanted is simply 1:#columns.