• hence it has to be the start of the statement,
• hence the line with the sub statement in our case
• not the "function name", since it could be in another line

Sorry, I don't think there is any room left to argue.

edit

If you type v you'll see something like

  DB<1> n
main::tst(caller_subline.pl:24):                               do { A 
+=> 1, B =>
main::tst(caller_subline.pl:25):                             );

  DB<1> v
21
22
23         print_calling_line( __LINE__,
24==>                          do { A => 1, B => 2, C => 3 }
25                           );
26
27         print_calling_line( __LINE__,
28:                            do { A => 1, B => 2, C => 3 },
29:                            do { A => 1, B => 2, C => 3 ,
30:                                   do { A => 1, B => 2, C => 3 }

  DB<1> n
Called from line 23; caller() reports line 24
main::tst(caller_subline.pl:29):                               do { A 
+=> 1, B =>
[download]

hence the line with the sub statement in our case

In the OP's case, yes.

I was talking about function calls in general. And for the case of displaying line numbers to a sentient being (which the OP did mention).

For a debugger, yes, the line of number of start of the function call is needed to be able to display the function call correctly.

Update: Though, for a GUI debugger, highlighting the line with the closing ) as long as there wasn't another function call on the same line.

Certainly, the least ambiguous is to have the line number with the start of the call (either the function name or the sub).

My main point of my previous post still remains: Anything line "in the middle" of the call is too ambiguous, therefore not acceptable.