Can you run the DMLtest function with different replicate group sizes? For me it fails with the following output:

Smoothing ...
Estimating dispersion for each CpG site, this will take a while ...
Error in x1 * wt1 : non-conformable arrays

if called with groups of 3 and 5 replicates.

It might not be the best idea to have different sample sizes in the first place but would also not be beneficial to exclude this additional information in the variation analysis.

Thank you for your help,

Edit: btw. with groups of 3 replicates each it does not fail