SuanShu should give similar results to R. They should not differ much. Maybe you want to check if you are calling the SuanShu classes properly, e.g., selecting the proper trend types (there are 5 trend types).
The reason why SuanShu does not give same results as R is because R does not give “accurate” results in the first place. R’s source code hard codes only a few critical values. SuanShu, on the other hand, reproduce the whole ADF distribution.
To get accurate distribution in SuanShu, you will need to supply the correct argument, e.g., enough simulation and discretization steps. Maybe you want to change the way you construct the ADF distribution table to see if this gives you results closer to what you are looking for.
Or, you can post your code (or PM me), I will take a look.