The WTA1 filter, which is used for 44.1 to 384k, has no change at all whilst going from 44.1 to 48k - it's merely the same filter run at a slightly different rate. But when you double the sample rate, then WTA1 goes into a double mode - so it's the same characteristic, but half the rate, as the input rate is doubled. In effect. 20 kHz becomes 40 kHz for 44.1 to 88.2.
For 705/768 WTA1 is in pass through mode, so it's not filtered. After WTA1 you have two choices - either WTA2 (which is a linear phase symmetrical WTA filter that takes you from 16FS (786 kHz) to 256 FS (12,288 kHz or 12.288 MHz) - the white and green filter - or an IIR filter. This 3 stage IIR filter either starts at 256FS (white or green) or takes over at 16FS to replace WTA2 (orange or red). The time constants for this filter is adjustable (about 240 kHz or 60 kHz) and this gives us the high frequency filter (white/orange off, green/red on). The intent on this filter is to remove the ADC distortion and noise that plagues HD recordings.
The HF filter is easily measurable - its about -1dB at 20kHz. The 16FS or 256FS is NOT measurable using standard tests, as it only affects above 352 kHz, and its subtle. It was a big surprise to me however, in that how audible the 16FS/256FS WTA2 filter actually is.
As too the Stereophile frequency response test this is actually incorrect. Qutest certainly does not drop like a stone above 19.5 kHz. This is because the WTA1 latency is so large, you can't use AP standard FR tests, as the latency makes the AP record the incorrect output. The actual frequency is at 21,990 Hz for 44.1 kHz. The noise plot is correct, as this test is latency independent.