I would add that API 650 says in 7.3.5 "Testing of the Shell"
"If water is available for testing the shell, the tank shall be filled with water as follows:
(1) to the maximum design liquid level, H;
(2) for a tank with a tight roof, to 50 mm (2 in.) above the weld connecting the roof plate or compression bar to the top angle or shell;
(3) to a level lower than that specified in Subitem 1 or 2 when restricted by overflows, an internal floating roof, or other freeboard by agreement between the Purchaser and the Manufacturer,
or
4) to a level of seawater producing a bottom of shell hoop stress equal to that produced by a full-height fresh water test."
IMO, this means also the test level can be different from design level.
Also IMO, strictly speaking, this isn't in harmony with 5.6.3.2 where both design shell thickness and the hydrostatic test shell thickness are required to be calculated based on the same H= design liquid level.
In case Appendix F is applicable, see also F.4.4 "When the entire tank is completed, it shall be filled with water to the top angle or the design liquid level, and the design internal air pressure shall be applied to the enclosed space above the water level and held for 15 minutes. [...]".
I guess API 650 would define more accurate "product design level" and "hydrotest level" as two notions that must be separately evaluated together with Purchaser.
However, for both cases (design for product and hydrotest) is clear there is no requirement to consider H as the top shell, even this is a common practice for "hydrotest".