I think its fine except the variable naming makes it a bit hard to see what’s happening. And the odd comment helps. Personally I think a recursive function like this is a good fit for the problem. Perhaps like this:

defmodule Roman do
  @roman_to_decimal_mappings [M: 1000, CM: 900, D: 500, CD: 400, C: 100, XC: 90, L: 50, XL: 40, X: 10, IX: 9, V: 5, IV: 4, I: 1]

  @doc """
  Convert the number to a roman number.
  """
  @spec numerals(pos_integer) :: String.t()
  def numerals(number) do
    numeral([], number, @roman_to_decimal_mappings)
  end

  # While the number is >= the decimal value at the head of the mappings,
  # accumulate the Roman numeral and decrement the number
  defp numeral(roman_acc, number, [{roman_numeral, decimal_value}| _t] = roman_to_decimal_mappings) 
      when number >= decimal_value do
    numeral([roman_numeral | roman_acc], number - decimal_value, roman_to_decimal_mappings)
  end

  # When the number is less than the decimal value at the head of the
  # mapping list, move to the next mapping pair
  defp numeral(roman_acc, number, [_h | remaining_mappings]) do
    numeral(roman_acc, number, remaining_mappings)
  end

  # And when the mapping list is exhausted, reverse the accumulated
  # Roman numerals and join then
  defp numeral(roman_acc, _n, [] = _remaining_mappings) do
    roman_acc
      |> Enum.reverse
      |> Enum.map_join("", &to_string/1)
  end
end