In answer to the original question:
Yes, there are different definitions of packing factor, and the one used in SAM is different from the one used in the document referenced above (
www.nrel.gov/docs/fy11osti/49417.pdf) and both are different from how I've seen it stated elsewhere, such as this DOE glossary:
www1.eere.energy.gov/solar/solar_glossary.html#P
In any case, the geometry depicted in the diagram you referenced is still useful in determining an appropriate packing factor.
Assuming the SAM definition of packing:
With flat (tilt = 0) modules, using a unity (1) packing factor gives only the area occupied by modules, so you will want to increase it to accommodate service access, inverter equipment, etc.
With tilted modules, the spacing chosen will determine when a module will become shaded by the adjoining module. This depends on latitude and tilt.
To get packing factor from row-to-row spacing:
p = (d / l) cos t, where
p = packing
d = row-to-row spacing
l = module width
t = tilt from horizontal.
Again, this only accounts for modules (and the spaces between rows), so increase somewhat for service access, inverter equipment, etc.