Potential bug with indirect left-recursion

I've isolated a 3-rule pattern, where I'd expect the parser to succeed for a given input, but it fails instead. The sample grammar:

RuleA ::= RuleB.
RuleB ::= IDENTIFIER
        | RuleC '.' IDENTIFIER
        .
RuleC ::= RuleB
        | RuleA
        .

The implementation:

import scala.util.parsing.combinator._
import scala.util.parsing.combinator.syntactical.StandardTokenParsers

object SimpleParser extends StandardTokenParsers with PackratParsers {
  lexical.delimiters ++= List(".", "$")

  lazy val ruleA: PackratParser[String] = ruleB <~ "$"
  lazy val ruleB: PackratParser[String] = (ident
                                       ||| ruleC ~> "." ~> ident)
  lazy val ruleC: PackratParser[String] = (ruleB
                                       ||| ruleA)

  def main(args: Array[String]) = {
    println(ruleA(new PackratReader(new lexical.Scanner("x.x$"))))
  }
}

It fails for input x.x$, telling me that it expects a $ instead of the ..

For me, this seems to be a problem with how indirect left-recursion is handled. I'm not sure if the original algorithm is incapable of handling this pattern, or this is an implementation bug.

Edit:
I've accidentally used | (first matching alt.) instead of ||| (longest matching alt.), I've fixed that in the code, but doesn't change the outcome.

Run the provided SimpleParser example with PackratParsers and the input x.x$ to reproduce the failure. Read the indirect left-recursion handling and compare it with the linked packrat algorithm; done means establishing whether this is an implementation bug or a limitation of the algorithm and recording the resulting behavior.