At this time, $1 substitution works by parsing a command line argument as a
number and pasting that number token into the script.  @1 turns it into a string
literal token.  There is no way to paste plain old characters, to do this:
   probe syscall.$1 { }
So, let's change things:

@1 would be redefined/clarified to paste a double-quoted (and internally
\-escaped) sequence of characters; $1 would be redefined to paste the command
line arguments verbatim.  In each case, these would paste characters right
into the lexer input stream.  (Numeric parsing for $N would be left to the
ordinary tokenizer.)

Existing scripts that use $1/@1 should actually be unaffected by this.