I should mention that the given structures are quite ok for short precursors (such as found in animals). For long sequences of more than 100bp you should not trust them. miRBase always shows hairpins even when the loops are very large.
Considering this, I would suggest to download the precursor sequence and mature sequence. Fold the pre-miR sequence on yur own with your favourite program (RNAfold, mfold/unafold, etc.). As miR and miR* are highly complementary I am pretty much sure that you will obtain the required stem in this section without a constraint for folding.