Breaking RSA on XMEGA - Plotting Bandwidth-Limited Data

In the notebook where we try to demonstrate an attack against an RSA implementation(sca202) I managed to successfully recover the 16-bit key via SPA using the SAD method, but when I try the attack using a low-pass butter-worth filter on the trace no matter how many different setups I 've tried the closest I got to recovering the key was 0x8053 (actual key : 0x8AB0).

When I plot the filtered trace I get something like this:

Here is a snapshot where I tweak the maximum threshold :

If someone has also stumbled on this one, it would be deeply appreciated to share his/her solution to this one!