My ANTLR grammar with optional token fails on simple statements

Question

I use ANTLR 2.7 to create a grammar to parse simple expressions like

1
2%
(3%)
(4)
((5))
((6%))

It's simple numbers which can have paranthesis and optional percent characters

My grammar looks like this:

class MyParser extends Parser;
options {
  k=2;
  buildAST = true;
}

parexpr
  : variable
  | LPAREN^ parexpr RPAREN!
  ;

variable
  : (NUMBER PCT) => NUMBER PCT^
  | NUMBER
  ;

class MyLexer extends Lexer;
options {
  k=6;
}
LPAREN:  '(';
RPAREN:  ')';
DOT: '.';
PCT: '%';
protected
DIGIT:  '0'..'9';
NUMBER: (DIGIT)+ (DOT (DIGIT)+)*;

I can successfully parse e.g.

(1)
(2%)

But it fails to parse a single number:

1

If I remove either | LPAREN^ parexpr RPAREN! or : (NUMBER PCT) => NUMBER PCT^ then I can parse a single number (but obviously not the other expressions).

I do not understand why I cannot parse a single number with this grammar..?

Note that I am working on a project that was implemented with this version a long time ago, and I'm just extending the existing things.