Using any kind of lowpass filter in general should introduce some delay, so the results you are seeing are to be expected. That said, if you wish to compensate for the delay, you could first design an equivalent butterworth filter using the " butter " function, and then pass in this filter with some additional parameters into the " grpdelay " function to compute the group delay.
Then you could just apply this delay to the final dataset to line up the plots of the two signals.
