6 months late. I found your question interesting, and so decided to take a deeper dive for my own sake. Not sure if this remains of any help to you at this stage.

Q1 : As stated in the UMH post you linked, a building model's surrounding grade/ground in EnergyPlus is always Z=0. This postulate is not only key for calculating atmospheric variations (e.g. based on the elevation of a wall, or that of an air node), but equally for solar calculations (e.g. ground plane/reflections), for instance. So the Z origin of typical mid-height spaces of a 22 storey building (e.g. 3m floor-to-floor height) should be ~33m. Lifting Z origins by a further 55m would make sense if instead modelling the 29th floor of some other 58-storey building. I'd initially argue that this is not a reasonable means to "account for the elevation difference" between weather station vs building location ... see below.

Q2 : "Best practice" can be broken down into multiple, in/dependent steps - depending on the circumstances. Let's start with most the basic EnergyPlus recommendation when one chooses to model a typical "middle floor" to represent many (e.g. 20 out of 22), inter alia:

Since exterior convection coefficients vary with elevation, locate the typical middle floor zones mid-height between the lowest and highest middle floors to be modeled.

Easy peasy. EnergyPlus adjusts zone, surface and air node temperatures at runtime based on their height (vs building ground elevation). As mentioned in the link you provided, the building ground elevation here is assumed to be that of the weather station, taken from the EPW file (not Site:Location elevation). EnergyPlus considers by default a standard WMO weather station temperature sensor height of 1.5m and a standard -0.0065 K/m temperature lapse rate (or what EnergyPlus calls an air temperature gradient coefficient). If the weather station temperature sensor were to read 15°C, then in theory the adjusted air temperature at the base of your building model should be 14.7°C, while dropping to 14.2°C near the roof. We're therefore looking at a 0.3K discrepancy.

Is such a discrepancy significant for natural ventilation assessments? Some may argue "yes". Then again, if outside air temperature acts as a control threshold ("close windows below X °C"), I'd argue that a 0.3K discrepancy could safely be ignored. A more relevant question IMHO is whether this 0.3K is in fact realistic. If the building is located in a suburban or urban setting, then one can easily anticipate warmer conditions (than the weather station) and reasonably ignore the 0.3K discrepancy altogether. So it depends ...

If you were adamant on correcting a 0.3K discrepancy, you have 3 options IMO:

1. adapt relevant EPW entries (e.g. DBT, WBT, Pa);
2. tweak the air temperature gradient coefficient; or
3. lift the building by approximately 50m (i.e. 55m - 3.7m - 1.5m), as suggested in your first question.

• Option 1 ...