We are counting on the xUnit "Publish xUnit test report" post-build step to properly set the build status. The Failed Tests section has 4 inputs that let you set thresholds to set the build to either UNSTABLE or FAILED based on whether the Total number of Failed tests, or total number of New failures, exceeds the specified thresholds.
There is something broken in the test of the number of failures against the specified threshold. A recent build has the number of actual failures < the specified threshold for Total, and the number of New failures (1) < the specified threshold for New (5). It set the build to failed anyway.
It appears that whenever the number of New failed tests is positive, that is reported as a build failure even when that number is less than the specified threshold.