A note on a difficulty inherent in estimating reliability from stress–strength relationships