Spike-noise in recorded track elevation while paused

I've unfortunately switched from a phone with pressure sensor to one without. Progress? No. After experimenting a long time with GPS "optimize GPS values" and filtering:( no | light | medium | heavy), Even with typically 38 satellites and 2m horizontal accuracy, I switched to "replace GPS values". This generally results in less vertical noise/ jitter in the recorded track elevation. However while the track recording is paused I often notice really bad elevations, today being an example of the worst.

track recording https://link.locusmap.app/t/maoar4

Here is the LM4 (Android) elevation because the web planner elevation doesn't show negative values for some reason (bug). I paused at 2km (half way) and clearly/ logically the elevation should remain close to 60m while stationary.

Here is the GPS accuracy. +/-22m of positional jitter. Why is this level of jitter sufficient to cause a huge change in calculated elevation given that the values are derived from SRTM data & GPS elevation is ignored? "replace GPS values" is set. The land around this pause location is relative flat, so even moving the position +/-200m will not result in a significant change in elevation, and certainly not -40m!!

The notional position when paused inside building seems quite reasonable: