Hi,
I did quite a lot of test with processing pictures before using them in Photoscan. I am modelling caves, so the problem of light is always present for me.
What you can do depend a bit of the quality of the picture you first took. For example, if it's noisy because your ISO sensitivity is set to high, you might exaggerate that noise in lightroom when post-processing it.
When postprocessing the pictures, I start correcting the exposures. And as David Cockey said, I'm also using the Chromatic Aberration correction and vignetting correction available for my lenses (turning off the distortion correction).
Then, I play with Highlight and Shadows recovering the overexposed and the underexposed areas. I always use the same parameter for the whole set of pictures. Changing highlights/shadows can improve the geometry of the 3D scene when those areas are really under/overexposed (meaning areas of full white or black pixels). If they are just a bit dark or a bit light, it seems that it will have no (or almost no) effect on the geometry of the model. But editing those parameters can still be interesting to get a more uniform light on the texture (less contrasted too), if it is what you want.
But you should not use filters on the pictures like "noise reduction", "sharpness", etc... It changes the pattern of the picture and might decrease the quality of the model. At least that is what I saw from the few test I did. I would not even push the contrast higher, which will exaggerate the noise.
To get the best model AND best textures, I should try to edit two set of pictures. To build the geometry with a first set, then texture the model using a second set of the same pictures, which were specially edited to look good. Anybody tried it already ?