I think you should use fully defined medium which is a medium that you know exactly the concentration of all its components. So you should avoid for instance yeast extract, peptone, tryptone ... etc. Instead you should use ammonium chloride, ammonium sulfate, ....... etc.
You can lump reactions if they are in a sequence, e.g. A -> B -> C -> D is equivalent to writing A -> D from a flux analysis point of view. This is a simple situation. A more general case is when the metabolic pathway between metabolites (A1 + A2 + .. + An) -> (D1 + D2 + .. Dm) is closed, meaning there are no external fluxes entering or exiting the network between all (Ai) and (Dj). Finding all these situations using software might be possible, albeit not a trivial problem.
You can lump linear pathways like in the A->D example or whole metabolisms into glucose + O2 + NH3 + S + Pi -> CO2 + H2O + biomass.
One should take care that with the lumping there will also be significant loss of information regarding the activity of each enzyme that is included in the original model. But I guess you can make the lumping consistent with the results you want to achieve.
You can select the grade of lumping according, first the objecteves that you are pursuing, and second, according to the degree of specific information that is available in relation to your biological system. It is not the same, to formulate a stoichiometric reaction to control or monitoring a bioprocess, that formulate a model to discern between alternative mechanisms hypothesis. As says Averine, you could move from a global reaction, to a very structured model, passing through a semi-structured model.
A god idea, could be first define three blocks: a) catabolism; b) anabolism (biomass synthesis); and energetics (respiration). There after you could develop in more details the block containing the molecules of your interest. You should see, if you consider in addition reactions dealing with primary (associated to growth), or secondary (independent of growtj), or mixtured (partially dependent of growth) metabolites of technological interest.
Comments of Tarek are also right. However, sometimes the available data come from experiments that used yeast extract, peptone, etc. Even in this case, you could make some applications if your consider the global composition of these substances, and added them as ancyllary sources of nitrogen, carbón, etc.
As general orientation, I recommend to read the paper of Noorman, Heijnen, and Luyben. Biotechnology Bioengeneering (1991), 38: 603-618 and the text-book of Roels, Kinetics and Bioenergetics in Biotechnology. Elsevier. Best