These are internal files generated during data processing for tracking and reproducible analysis. They can be identical or different depending on which options you have chosen - although the file names could be misleading in some cases
If you are referring to the “Linear model with covariate adjustment”, the covariates are included in the multiple linear regression model used to calculate the p-values and effect sizes for each metabolite with respect to the metadata of interest. They are not used to directly modify the data.
Is it supossed I can use the output file of normalized data.csv for future analysis to avoid the normalization step? However, the re-used normalized data.csv (obviously not normalized in a second step) leads to a different normalization profile, so I interpret there is something new being introduced that I do not control.
Should I upload my original, raw data at each step, and do not reuse data provided back from Metaboanalyst?