Having read the previous 3 posts I feel compelled to write that, having bought deserter's old obelisk 1s from johnacurtis once John discovered they were too bass heavy in his listening room and then being dismayed to find there was no output from the tweeters and thinking I'd have to return them and ask John for my money back, I came across a post from Linus saying what he's just written above and, having a NAP180 in my annexe system swapped it for the olive 250 I was using in the main system and Bingo!
Apparently the power supply regulator boards in the 250 and 135s contain current limiting circuitry that prevents the amps from blowing up into dangerously low impedances. Trouble is the current limiting has an unintended frequency dependent character that restricts current more as frequency rises. The NAP 250 made Naim's reputation on its ability to drive Linn Isobariks (less than 2 ohms at around 50 hz) but faced with similar impedance at high frequencies it may not be able to deliver the required current (depending on sample variation - I must have got an extreme case).
Going from 250 to 180 was like having an on/off switch on the tweeters. With the 250 the switch was off, with the 180 the tweeters were switched back on again, it was that pronounced!
I understand the Obs 2s are more forgiving in that respect, and your Witch Hat modified 135s may be immune from that problem anyway, but I'd be inclined to give John Burns at Pear Audio a call to make certain, I'm sure he'll know the answer!