Add note to user_guide.md

docs/user_guide.md

@@ -33,6 +33,12 @@ in ipolates_mod. The interpolation method is chosen via the first
 argument of these routines (variable IP). Sub-options are set via the
 IPOPT array.
 
+It should be noted that some routines may behave poorly or unpredictably when
+using 4-byte reals (-DBUILD_4=ON). For instance, there is an ATAN2 function
+used for polar stereo grids where for certain grids/coordinates, floating point
+differences between 4-byte output values (~1e-7) can be amplified into
+noticeable differences in output field values.
+
 ### Bilinear Interpolation
 
 Bilinear interpolation is chosen by setting IP=0.