I'm not going to turn this thread into a physics class, but the basic idea is as follows

• light can be considered as a set of beams in loads of directions, beams can be considered as a stream of photons (bullets) fired towards a surface
  (yeah I know, there are wave interpretations too, and so on, I passed my Quantum Electro Dynamics exams. Not relevant here.)
• surface can be considered as a large field of microfacets, all with different directions. The more those directions are similar, the smoother the surface.
• when a photon hits a microfacet, there is a change A% it gets reflected, B% it gets diffused, C% it gets involved in translucency, D% it passes through and gets refracted, and 1- (A+B+C+D) it dies because of absorbsion (and heats the object).
  You can also say: A% of the incoming light gets reflected, and so on.
• as a consequence, A+B+C+D should be < 100% when representing real life stuff.
  unless of course you want to show response to hidden lights, or want the object to behave light a neon-sign, etc.
• The A% has to be divided over reflection and specular, the specular itself over specular and alt_specular. More of one means less of the other, they are all faces of the same medal (in nature!)
• The B% has to be divided over Diffuse and Alt_Diffuse, and so on.
• on top of that, surface / geometrical considerations have to be taken into account: light is not diffused, reflected etc in all directions evenly. Some model ("Lambert")for this is build into Diffuse and Specular, other models (Blinn, Fresnel, ...) are available through the Alt channels which is exactly the purpose of these.
• But since Alt_Diffuse etc just passes through what it gets, some shader builders create their own shading system by creating their own node tree with really everything in it, and route the result throught the Alt_Diffuse channel.
  Fine, with its own advantages and disadvantages.
  Fun test: set all channels to 0/black and just plug in a plain color node into Alt_Diffuse. Render. See what happens without handling the geometrics.
• Poser cheats in a lot of ways, as it's intended to get artistic results, not photoreal. So it lets you have 100% Diffuse + 100% Specular, or 100% Specular + 100% Reflection, and so on. Great, but not for photorealism.
  For that purpose, a first step is to ensure that the sum of values does not exceed 100%. The basic principle behind this is the law of conservation of energy. Physics-based renerers like LuxRender have this incorporated to a deep extent. Poser of course has not, but we can come close manually.
• Poser cheats even in more ways. So will specular respond to direct light only and not to environmental light (IDL), reflection responds to objects around and not to light at all, and so on. Great for render speed, not for physical accuracy
• whether lights emit and surfaced return lume, or watt per square inch or whatever is not relevant. No one knows what 100%-intensity exactly means for a Poser light.

whatever the internal details, the renderer essentially derives results per light per channel, and adds these up to the final image. You can do the same by hand, when you get images per light per channel you can add them as Photoshop layers, using the Screen blending mode. Doing this really properly requires some extra gamma magic, by the way (do have image gamma, but no render gamma. Do gamma at the end, in Photoshop).
When doing render passes, you can render shadows seperately, and use Multiply in Photoshop. Life is simple.

Since Poser-internally all lights, materials and render (sub)results are handled in 16bit colorformat, the first clipping is the result of exceeding the internal maximum value. Everything over 100% is just overlit. The extra bits create no extra room in brightness handling capacity, as white just reads RGB=(65535, 65535, 65535) instead of (255, 255, 255). Black always reads (0,0,0) and 16bit just gives more details inbetween.

The second reduction comes indeed from mapping that 16bit color to an 8bit color, required for most image formats and for showing things on screen. But as white remains white and black remains black, there will be mainly loss of detail and hardly brightness clipping involved.

How to measure the brightness of an area? It's the V (0..100%) in HSV. Open the image in Photoshop, open the Info panel, use the eyedropper / colorpicker and read out the value. Or whatever works in GIMP, Paintshop, Canvas, etc.

have fun, thanks for the attention, final extended and illustrated tutorials coming soon.

- - - - - 

Usually I'm wrong. But to be effective and efficient, I don't need to be correct or accurate.

visit www.aRtBeeWeb.nl (works) or Missing Manuals (tutorials & reviews) - both need an update though