Constant-density fluid "expands" as it is emptied from a tank
 

I am running a transient, multiphase (VOF, 3 phases), axisymmetric simulation of a fluid being extracted from a tank through an outlet placed at the bottom. I am initializing the tank with two layers of the same fluid (constant density, Herschel-Bulkley viscosity). Since they are the same fluid, surface tension between them is set to 0. The third phase is air.

These are the registers used to separate the two fluid instances:

[Cell register for fluid on top]
[Cell register for fluid on bottom]

These are the tank states at the start and end of simulation. The two fluids are separated by the gray/black line:

Tank at t=0.
Tank at t=6.5s.

As you can see from the numbers at the top, the volume of the top fluid increases from 10,800L to 13,000L between the two states, even though it should stay the same. The volume is derived from the use of CFD-Posts 'areaAve' function multiplied by a constant. The function's input is the corresponding fluid's volume fraction. I have confirmed it is not simply an issue of incorrect 'areaAve' usage. The top paste at t=0 occupies less pixels than at t=6.5.

What could be going wrong with the simulation that causes this fluid's volume to change?

Can anyone help with this? I've been trying to find the fix for 3 days now but still can't figure out what's going wrong. I tried changing VOF settings, different schemes, different georeconstruction methods, time steps, meshing, etc but the fluid still gets larger.


EDIT: Mass is conserved when running the problem in "Planar" (2D) mode vs "axisymmetric". But since I am simulating a cylinder, I need the problem to be axisymmetric.