On the biases affecting water ages inferred from isotopic data