Strange measurement

I'm trying to measure the loaded voltage of a pair of headphones. I take on channel from headphones out, split it and loop it back to input 3 (RME UCX sound card). The other split goes to the headphones left channel. I know that the headphone output of that sound card has a 25 ohm output impedance. The headphones have a nominal 50 ohm input impedance which should vary with frequency. In fact, when I measure using a multimeter, I can see a maybe 5 dB effect when sending out a sine wave of from 100 to 10000 Hz. This is what I expect. When I measure simultaneously on ch 3 input using REW, I get a perfectly flat curve that does not show the frequency dependent output voltage I know is there. I have been over it forward and backward and it does not make sense. I'm using the "instrument" setting on the ch3 input, which makes that input channel very high impedance so it does not have any effect. I have not tried to use any calibration files or anything like that.

The upshot is I can measure frequency dependent changes on input 3 using a multimeter but see no such effect when measuring transfer function in REW. Ruler flat measurement! Can't figure it out.

False alarm I guess, I traced it down to the multimeter! Apparently my several multimeters I tested with are not good above maybe 1 kHz. When I plug the multimeters directly into a signal generator, I also see the drop-off in voltage at higher frequencies up to 10 kHz on the (Fluke) multimeter and another meter. When I tested on an old, and much better meter, it was flat as you'd expect.

The only thing I can't figure is WHY there is so little impedance loading. Seems like there ought to be more than what I'm seeing in the REW measurement. Anyways, I guess not an REW issue!