Not sure why I'm responding to a one year old thread, but this confused me initially too, so I guess I'll clarify for anyone with this question in the future.
(I was confused about how the mystery mark could exist, as a median ATAR of 70 suggests a minimum ATAR of 40 if ATAR scores are evenly distributed)
According to the UAC Scaling Report (2019), the ATAR actually accounts for the whole HSC age population (number determined by ABS statistics). Hence, the ATAR is a percentile of your performance in the whole HSC age cohort, rather than just the people who receive an ATAR. UAC uses a model (cubic spline model? not really sure how they came up with this) to estimate your percentile even though not all HSC-age people attain an ATAR or even complete the HSC. The aim of the model is to determine your percentile amongst this whole population, and not just those who get an ATAR.
Effectively, the model states that all of the most capable students (i.e. 99.95ers) attain an ATAR, so it assumes that the top 0.05% of ATAR-eligible candidates all attained an ATAR. However, as you go down, the proportion of non-ATAR students who could have received a certain ATAR increases, so in general, less people will lie on a lower ATAR than a higher ATAR. For example in 2019, there were 923 students with an ATAR from 99.00 to 99.95, 900 students with an ATAR from 75.00 to 75.95 and 700 students with an ATAR from 49.00 to 49.95.
tldr: the reason why the median ATAR is 70, when the ATAR is a percentile, is because ATAR accounts for the whole HSC age population, not just those who achieve an ATAR.