A test case reused the stub left by the case above it, so it failed on that stub's data, not on the rule it named. Break the rule and the test stays green, because the stub still makes it fail.
Reset every stub a case relies on. A leftover stub makes the case fail for someone else's reason.
Make each negative case fail on one assert only. Then you know which rule it checks.
Make each negative case fail on one assert only. Then you know which rule it checks.