Based on this thread
and what Thomas Mueller pointed there, you might define types with the same signatures as methods whose addresses you want to obtain (for each overload). If you then declare the variables of those types and assign method pointers to them you will make sure that compiler chooses the right overload to your known variable type and moreover that it won't ignore them if they wouldn't be used anywhere in the code (some overloads might not get linked in your binary).
So based on his idea it might looks for the MessageDlgPosHelp
function overloads like this:
type
TMessageDlgPosHelp1 = function(const Msg: string; DlgType: TMsgDlgType;
Buttons: TMsgDlgButtons; HelpCtx: Longint; X, Y: Integer;
const HelpFileName: string): Integer;
TMessageDlgPosHelp2 = function(const Msg: string; DlgType: TMsgDlgType;
Buttons: TMsgDlgButtons; HelpCtx: Longint; X, Y: Integer;
const HelpFileName: string; DefaultButton: TMsgDlgBtn): Integer;
procedure TForm1.Button1Click(Sender: TObject);
var
MessageDlgPosHelp1: TMessageDlgPosHelp1;
MessageDlgPosHelp2: TMessageDlgPosHelp2;
begin
MessageDlgPosHelp1 := MessageDlgPosHelp;
MessageDlgPosHelp2 := MessageDlgPosHelp;
ShowMessage(Format('%p; %p', [@MessageDlgPosHelp1, @MessageDlgPosHelp2]));
end;