Dear sivastatic,

1. Appendix F refers to Design of Tanks subject to Small Internal Pressures. Under the internal pressure, the roof is subject to lift tendency (and a component is a lift-off tendency in shell) but the upper ring is also subject to a compression load (tries to collapse inside). That's why the title of F.5 is Required Compression Area at the Roof-to-Shell Junction.
Comparing to the general requirements of API 650 in 5.2.2 it appears that F5 formula address the roof-to-shell junction under 5.2.2 Load Combinations point a) Fluid and Internal Pressure: DL + F + Pi (heavy fluid is not in contact with the junction,so equally the case is DL + Pi).

F5 formula does not differentiate the cases in which you have or you haven't roof rafters attached to the upper shell and/or to compression bar. F5 formula is a simple one counting a dome or cone membrane attached to a cylinder.

2. In my engineering understanding, one still need to demonstrate that the roof and the roof junction have been checked to all points of 5.2.2. This is not usually happening in practice because also API has been changed several times and the loads definition was in Appendix R previously (and the designer had the alternative to say that the Appendix was not mentioned as mandatory). But now is a requirement of basic API 650, probably the Clients will strongly ask for. On my side, I cannot argue why there is no need to comply with all points of 5.2.2.

3. A part of 5.2.2 is to count the Gravity Loads points e1 and e2. Now the effect is an axial force down in the upper shell but also a tension in the upper ring. There is no indication how could be counted- because e1 and e2 are explicitly reffed under the titles Self-Supporting [...]. Let's say that under your efforts to comply with e1 and e2, you are able to perform a FEA analysis including the junction and the rafters connection, hence you are able to demonstrate the mechanical tensions in junction are within allowable- that solved this point in analysis. In case you haven't such tool, one alternative is to count the necessary area similarly with the case described in F5; the difference is that the Dead Load is not acting against pressure but it is rather the load generating the mechanical tension stress in junction. This approach (that does not make difference you have or you haven't roof rafters, exactly as F5 approach) leads to a formula similar with 5.10.6.2, regardless the roof is defined as self-sustained or not.

4. In my opinion, including 5.10.6.2 under the title 5.10.6 Self-Supporting Dome and Umbrella Roofs is not the best choice of API 650. The original intention of API was to define the self- sustained dome type only as a membrane with spherical shape sustained by the upper shell. Having rafters you may say it is not your case, your Client may say it is, etc. But the engineering logic remains as I tried to describe in my point 3. And note that by API calculation anyway you are not addressing at all the local buckling in upper shell and roof, so a prudent approach is to consider 5.10.6.2 anyway, even it gives you the need to increase that area vs. what is mandatory, i.e. formulas in F5. If I were you, I would consider it. If I were your Client I would ask for compliance with all 5.2.2 Load Combinations checking.

As you said, you must decide.