FSDS in the history file should have the same values as the shortwave radiation you are using as your forcing for a single-point simulation if you are running the model at a time-step that corresponds to your forcing data. E.g., if your forcing data is at half-hourly temporal resolution and the model is run at a half-hourly timestep (dtime=1800), then FSDS should have the same values as the forcing data.
Maybe your forcing data is being interpolated by the datm?
The "tintalgo" for the forcing data should be set to "nearest" in your datm_in.
E.g., here is a typical datm_in for the single-point simulations we run here (this is for the tower site "IT-PT1"):
&shr_strdata_nml
datamode = "CLMNCEP"
domainfile = "/glade/work/oleson/PLUMBER2/input_files/IT-PT1/domain.lnd.fv0.9x1.25_IT-PT1.nc"
dtlimit = 1.5, 1.5, 1.5
fillalgo = "nn", "nn", "nn"
fillmask = "nomask", "nomask", "nomask"
fillread = "NOT_SET", "NOT_SET", "NOT_SET"
fillwrite = "NOT_SET", "NOT_SET", "NOT_SET"
mapalgo = "nn", "nn", "nn"
mapmask = "nomask", "nomask", "nomask"
mapread = "NOT_SET", "NOT_SET", "NOT_SET"
mapwrite = "NOT_SET", "NOT_SET", "NOT_SET"
readmode = "single", "single", "single"
streams = "datm.streams.txt.CLM1PT.CLM_USRDAT 2002 2002 2004",
"datm.streams.txt.presaero.clim_2000 1 2000 2000",
"datm.streams.txt.topo.observed 1 1 1"
taxmode = "extend", "extend", "extend"
tintalgo = "nearest", "linear", "lower"
vectors = "null"
If you still have problems, please attach your lnd_in, datm_in, and log files.