Amplitudes values multiplied by unknown constant (Y axis skewed)

P.S. Indeed, filtering before cleaning solves the problem. Or rather, I should say, ICA was performed. However, I did not immediately proceed to apply SSP projectors. Instead, I performed the bandpass filtering with the "Process the entire file at once" checked. Thus, the SSP projectors were applied in the same step as the filtering. Though, I don't understand why this would be functionally different from my original pipeline. It shouldn't matter if bandpass filtering is done before or after cleaning, no?

Here is the new ERP with bandpass 1-50 first, then ICA: