2 points immediately spring to mind (would right more but have to catch a bus)
1. when modelling anything in watercooling the restriction tends to be of the form a*x^2. This is the case with your coke bottle rad above. This applies to everything and if your modelling something always compare your complicated formula to an x^2 graph unless there is a huge discrepancy. The x^2 result will be more “right” and any difference is measuring error. This might not apply for really complicated flow scenarios like a jet impingement block or a pump but should give you a very close answer. T

2. A coke bottle is not a good radiator as the water just rushes in and out, without settling, which is bad reservoir design. This has two effects
a. Air does not filter out as easily as the water never settles (moves really slowly) so air can’t escape.
b. There is lower restriction as the inlet and outlet are coupled together in a flow pattern. This reduces restriction but means the flow is not settling.

Restriction of a good reservoir with the same sized pipes going in and out will always be equal to (for SI units)

P = V^2 *rho

Where:
P= the pressure drop
V = velocity of water in m/s in the pipes going in and out.
Rho = the density of water 1 kg/ litre (don’t start putting stupid numbers of significant figures in)