Do you have a link to the directional coupler(s) circuitry you are using?
If it's a true representation of the voltage of a properly matched 50 ohm system then power is the product of (E X E) / R, where R = 50 ohms. If simple diodes are being used to rectify and measure the RF amplitude be aware that there will be non-linearity as the lowest power levels until the detected voltage is above the Vf rating of the diode(S). There are 'compensation' circuits that can help with that for diode detection, but one first must look at the total dynamic range one wants to have and what acceptable accuracy tolerances you can live with.
The better RF SWR meters available these days can give very good performance using log amp detectors, with many popular designs based on the Analog Devices AD8307, good for a claimed 92 DB range and 0-500Mhz bandwidth. With a log amp the output voltage is a straight linear representation of wattage.
I built a RF milliwatt meter based on one of those chips and the accuracy has been very impressive and tracks +/- 1 db difference between from about -77dbm to +15dbm when driven by a HP signal generator with an accurate selectable decade attenuator output.