I can look into this. @rrm : if you have a single parameter of interest (i.e. your problem is one-dimensional) it is recommended to use for FC the StandardHypoTestInvDemo,C , see
(notice also the update link to the doc, you are using the old , obsolete documentation) https://root.cern.ch/doc/master/StandardHypoTestInvDemo_8C.html
with the parameter for FeldmanCousins:
type = 0 (Frequentist calculator)
testStatType = 3 ( 2-sided profile likelihood test statistics: this is what defines FC)
useCLS = false ( the interval is computed using the CL(S+B) and not CL(S+B)/CL(B) as in the CLs case )