Struggling to understand how this rule is mutually left-recursive

[ethindp]

So, I've been writing my own version of the Ada grammar (for Ada 2022). One of the major left-recursive rules is name (Ada LRM, 4.1(2/5)). The original production is:

name ::= 
     direct_name | explicit_dereference
   | indexed_component | slice
   | selected_component | attribute_reference
   | type_conversion | function_call
   | character_literal | qualified_expression
   | generalized_reference | generalized_indexing
   | target_name

direct_name ::= identifier | operator_symbol

prefix ::= name | implicit_dereference

explicit_dereference ::= name.all

implicit_dereference ::= name

The Ada 2012 Grammar rewrites this as:

name
   : direct_name
   //| explicit_dereference
   | name DOT ALL
   | name '(' expression (',' expression)* ')' //indexed_component
   | name '(' discrete_range ')' //slice
   | name DOT selector_name //selected_component
   | name SQ attribute_designator //attribute_reference
   | type_conversion
   | name actual_parameter_part //function_call
   | character_literal
   | qualified_expression
   //| name                       //generalized_reference
   | name actual_parameter_part //generalized_indexing

   ;

However, this is missing a few extra productions, so I modified this to suit my grammar:

name:
        direct_name
        | name FULL_STOP ALL
        | name LEFT_PARENTHESIS expression (COMMA expression)* RIGHT_PARENTHESIS
        | name LEFT_PARENTHESIS discrete_range RIGHT_PARENTHESIS
        | name FULL_STOP selector_name
        | name APOSTROPHE attribute_designator
        | ((value_sequence | name) APOSTROPHE reduction_attribute_designator)
        | type_conversion
        | name actual_parameter_part
        | CHARACTER_LITERAL
        | qualified_expression
        | name actual_parameter_part
        | target_name
;

But Antlr still believes that this is mutually left-recursive. This rule (name) is the (only) rule that Antlr gives me when it generates Error 119. How else am I supposed to refactor this without risking breaking the syntax?