In the words of Ernest Rutherford, if your experiment needs statistics, you ought to have done a better experiment. 

Statistics is very important in comparing the nutritional composition of food samples. 

Thanks.

If you want to make comparisons between studied population groups on which you did your experiment YES of course.   

The nutritional composition of designed experimental diets, before doing an intervention, I don't think need to be compared with statistics, BUT the degree of adherence of your subjects to the prescribed diets during a trial, of course needs to be reported in statistics. Nutritional compositions of the diets reported need also statistics

There is no way you will use an experimental diet in a research without either simple or complex calculations. Whatever calculation you involve must be reported either descriptive or robust. For instance, if you are evaluating nutritional composition of under utilized vegetable in your country, the nutritional composition analysis ought to be done in duplicates or triplicates but at least you will need to calculate the mean and standard deviation of your results which is statistics. Application of appropriate Statistical tools are needed to make important conclusions in an experimental design.

Your question is not clear. If you wish to document result of analysis of a product, statistics of analysis may be required. If you plan to publish the results of a feeding experiment, the mean values should do. Think if statistics of analysis fit into your model, and there is a logical relation of statistics with the outcome.