I don't know exactly why the precision here is a problem, but if you increase the precision (something like -1.58989`50) you get a result of about 0.056. And if you use an exact value (e.g. -158989/100000) then you get an exact answer, which, when you apply N, also gives 0.056.
