Given the preceding discusion, my conclusions are sumarized here:

1. Current synthesis models are an optimun tool for obtain an aproximate idea of how star forming regions works, but unfortunatelly, they are not ideal for the interpretation of data obtained with high resolution telescopes.

2. Each model has their adventages and their limitations. I do not think that at this moment there are a really complete model. Read the papers (specially the limitations) and chosse the model that may better fit your requeriments (or ask to models makers, I sure that more of us will be happy to make extension of our codes).
   Do not use synthesis models like a black box.

   In particular, take care with the mass normalizations of the results of the code.

3. In the modeling, the main point that must be revised is the track interpolation scheme. All models produce simillar results, but the compute the isochrones in a simillar way. This problem, quoted by B. Tinsley, has not been considered in general (some exceptions are some works from Bertelli and other works by Brocatto) and there is almost no bibliography in the subject (neither a mention of caution).

4. In particullar, from my point of view, the use of tracks with twice mass-lost rates like the set published in 1994 by the Geneva group, must not be used since the actual isochrone computation may produce unphysical results that are more dramatic for larger mass-lost rates and these tracks are the ones that fit the worst with observations of individual stars. The results obtained with this set of tracks may be unrealistic (independent if it looks to fit the observations of integrated properties).

5. Other evolution channels like rotation, magnetic fields and binary systems must be also investigated in the context of star forming regions. Unfortunatelly, these three items are closely related and it is quite difficult isolate any of the problems.

6. In Nature, sampling is a natural effect, and it must be considered in the modeling. If your results are bizarre in some way, compute the mass transformed into stars to be sure if sampling may be relevant or not there.