I think it's possible that you maybe jumping to conclusions. IMHO, your post highlights the difficulty of testing the perception of sound. The fact is, that it's very difficult to isolate and test what we think we are actually testing and even experienced scientists sometimes get it wrong.
In your test for example there are a few points which, in my opinion, make it unlikely that you perceived what you think you did. 1. LPs are notoriously inaccurate and noisy at high frequencies, 2. Filters are inherently non-linear and produce audible artefacts well inside the pass band and 3. Of the energy produced by all a violin's harmonics, only 0.04% is above 20kHz.
It's virtually impossible to create an analogue filter operating in the hearing spectrum which does not introduce audible artefacts. Even the latest digital technology needs very careful programming and a lot of resources to implement a filter without audible artefacts. We also have to consider that with only 0.04% of a violin's energy being produced above 20kHz, that even with the very best and latest high tech equipment, it's still bordering on the impossible to record this energy above the noise floor of the recording environment and equipment. Let's say for the sake of argument that it is possible, we come to the last and most telling argument: Even if it is possible that someone can hear a signal above 20kHz, hearing a test tone at 20kHz is trivially easy compared to actually differentiating >20kHz signals from the noise floor and which comprise only 0.04% of the sound you are hearing.
G