Here's my analysis.
(**myint)(&k) -- cannot convert &k (type **int) to type **myint:
type **int and type **myint are unnamed pointer types and their pointer base types, type *int and type *myint, don't have identical underlying types.
If T (*int or *myint) is a pointer type literal, the corresponding underlying type is T itself.
(*myint)(k) -- can convert k (type *int) to type *myint:
type *int and type *myint are unnamed pointer types and their pointer base types, type int and type myint (type myint int), have identical underlying types.
If T (int) is a predeclared type, the corresponding underlying type is T itself. If T (myint) is neither a predeclared type or nor a type literal, T's underlying type is the underlying type of the type to which T refers in its type declaration (type myint int).
(myint)(*k) -- can convert *k (type int) to type myint:
type int and type myint have identical underlying types.
If T (int) is a predeclared type, the corresponding underlying type is T itself. If T (myint) is neither a predeclared type or nor a type literal, T's underlying type is the underlying type of the type to which T refers in its type declaration (type myint int).
Here's the underlying type example from the Types section revised to use integers and int pointers.
type T1 int
type T2 T1
type T3 *T1
type T4 T3
The underlying type of int, T1, and T2 is int. The underlying type of *T1, T3, and T4 is *T1.
References:
The Go Programming Language Specification
Conversions
Types
Properties of types and values
Type declarations
Predeclared identifiers
Pointer Type