ChAMP does not necessarily have to read iDAT files. Please check the vignette and documentation for the program. Essentially, the starting point is:

champ.process(fromIDAT = FALSE, ...)

I have processed methylation data from the TCGA and one does not even require ChAMP. You could simply read in all files and merge into a single data matrix of methylation beta (β) values. Then, your statistical comparisons would involve a Wilcoxon Signed Rank test between your groups of interest. You should add a 'difference in mean methylation' between each group as an extra cut-off, too (for example, mean(tumour) - mean(normal)).

Kevin