I see. The problem is that the original high pass filter is not being consistent with the amplitude, it actually raises the amplitude of some parts of the spectrum (!), and when you connect them in series the effect is multiplied. I haven't tried doing the same with [iemlib/hp1~], but I bet it will work perfectly well, since it behaves as a proper filter does: -3dB at the cut-off frequency, and no amplification of ANY part of the original spectrum.
Miller has a saying for this, when he finds a bug or limitation he doesn't like: you get what you pay for!
Nothing irritates me more than this quote, and it is disheartening to hear it from an open source developer... that is the motto (or rather mantra) of all executives that brush Linux and open source projects off. I understand the context he says this, but still...