I think you need to concentrate on the smooth wall simulation before trying a rough wall simulation, especially if the pressure loss is off by a factor of 4. The rough wall simulation should only add a marginal increase in pressure loss.

A value of nearly 1 ft for the roughness is another indication that something is a miss, assuming it is larger than the pipe diameter.

If you haven't already I suggest you create a 3 volume flow domain. You need to extend the inlet and outlet to be far away from pressure gradients (say about 6 elbow lengths away):

1. Extrude the elbow inlet and outlet using the Others->Extrude tool, with the Keep Operands enabled (to preserve your original faces) to produce 2 new volumes
2. Use the Face->Connect tool to connect all faces - it will only connect those faces that are near identical, i.e., the inlet and outlet faces for the elbow to the extended volumes
3. Create a new group of all 3 volumes
4. Assign the substance to the new group and set up your simulation as usual, e.g., reference velocity, boundary conditions. The only thing to watch is to make sure you do not assign a boundary condition to the 2 internal faces at the inlet and outlet to the elbow, respectively

Make sure the geometry is reasonably resolved by the mesh using the Conditions->Accuracy tool. Look at the Scalar Field->y+ on the walls. You want the majority of values to be in the range 30-300 otherwise your results will be compromised. You may have to use the Accuracy tool on the walls to get the y+ values within this range. y+ for a given flow configuration is directly proportional to the surface cell size.

To determine the pressure loss you need to create a new Result for the average pressure on a face as described in the comment "Create new Caedium Results". Then apply the pAverage monitor to the inlet and outlet faces of your elbow.

Assuming you are going to run a steady-state (default) simulation you want to see the residual monitor with all the residuals below 1e-3 and the pAverage monitors should converge to a single value with increasing time.

If you don't get within a factor of 2 for the pressure loss using this approach then there is no point trying the rough wall variant.

In answer to your questions:

The rough wall functions are configured by:

• Cs - Tuning parameter (0-1) - leave it at 0.5 unless your calibration implies otherwise
• Ks - Sand-grain roughness height, where smooth = 0 - units are as indicated, in your case ft

As a reference for sand-grain roughness related to measured surface roughness try:
"A Simple Algorithm to Relate Measured Surface Roughness to Equivalent Sand-grain Roughness"