I have always tried to use actual data whenever possible or refer to adjacent station data within physiographic area.  It has been many years since I used the SCS-CN method, so I am not familiar with its updates if any.  I am not a great fan, but if I were to use it, I would do what you are doing to try to validate or adjust to the local circumstances.  From memory, the SCS method relies primarily on soil types and expected response, but has little to do with vegetation type.  To generalize from a few NC USA studies, ET from grass cover may be 10 inches, hardwoods, 20 inches, pine 30 inches.  Accounting for landscape depressions or wetlands, geologic variances, hydrologic modifications, stream type and efficiency, etc. requires more than a basic model to address specific conditions.  So if you also have other gauging stations in physiographic area, compare their results with SCS Method to see if differences are consistent.  I am also unsure if SCS method is a model that should be used for daily, weekly records, but more for monthly or annual comparisons.  

For experimental watershed research, I would always measure flow and if catchment size, install flume or weir, use water level recorder.  The digital recording transducers today are much lower in cost than decades ago.  My researchgate thesis (1975) dealt with calibrating catchments for future experimentsl study.  For that type work, real measurements are used.

However if data is not available, sometimes we must consider or use the models available, and then try to validate as best as we can.  I think I would consult with a soil scientist, hydrologist and geologist as well as GIS specialist if there might be something missed, an issue with conditions and try to estimate reasons why my actual data of streamflow is consistently higher or lower, and by how much, how variable.  If you can find reasons for why the model reads higher or lower in your analysis area, it may be helpful to others.  

I have used the RUSLE model alot in my career, but in writing results, I always try to tell or disclose assumptions and indicate the calculations are estimates, applied at catchment or watershed scales with limits, useful primarily to to compare alternative actions, but not to be considered as real data, or with defined expectations relative to accuracy and precision.  

Your actual data indicate issues with the model in your area or specific catchment that need resolving for continued and best or most reliable use.  The challenge comes how to adjust the model in some logical way to better fit your analysis needs, or perhaps there is another model that better meets your intent.  A network of well located stream and rain guages, excellent soil maps, etc. can provide invaluable information fot the future and also be useful to validate, adjust or generate models.

i rambled on, have not used the SCS Method extensively and had issues to deal with when it was tried, but hopefully some of this will be helpful as you proceed.

if I have well understood the question, the reply is easy.

You have runoff in mm/day from scs-cn method, and discharge observations in m3/s. Simply multiply runoff by basin area and you will obtain data comparable with observed discharge (of course, you should adjust for the unit of measure)

regards, L

CN is a parameter, which relates the total rainfall and direct runoff. You should adjust the input data to output, resetting the average CN value of the basin, as a feeds back process.

i would like  thanks to all who answered my question.

@ Mr Luca Barocca as you told, i have been done in that way, and in order to convert surface volume to flow (time distribution), I converted SCS-CN calculated per day total volume in m3/sec (cumecs) by dividing it to seconds in a day (24*60*60) to cross check it with observed flow in m3/sec.

I have also done reverse calculation by converting observed flow in a day to quantify total volume passed to the most distant outlet to relate with corresponding SCS-CN calculated volume in a day. But calculated volume is exceeding 50% to 200% to observed discharge.

Generally, on account of factors such as catchment slope, length, surface roughness etc., it is not necessary that calculated one-day volume will be near to corresponding day observed discharge (equal to volume). I am interested to know that, comparatively, how much variation is acceptable for one-day calculated runoff volume to be passed from the outlet? or what else method to relate calculated volume with observed discharge?

Regard

S.S.Saini