For the accuracy needed to compare rocket propellants, I had to abandon every estimation method and rely exclusively on measured enthalpies of formation, checking with disconfidence if a published enthalpy had resulted from some software, and whether the compounds are in the proper state at the right temperature.
Without experimental data, the least bad method is indeed hand evaluation. This does make sense with your branched compounds without short nor multiple rings - but mind the two double bonds. Hand evaluation has some value only if you can refer to very similar compounds in the same state at the same temperature and apply minor corrections. Bond energies are already far less accurate.
Additive rules, preferably improved with nearby interactions rules, have the big drawback of neglecting the state and temperature of the compound - that's a superiority of hand estimation based on similar compounds.
Molecule modeling software (I have ArgusLab on Windows, Linux runs more varied software, all rely on the same Am1/Pm3/etc models) use to be >20kcal/mol wrong on such molecules, which is useless to my needs. Software generally computes an enthalpy of formation (...from the molecular elements, check it) for the gas at zero kelvin. Some software also estimates the heat capacity of a gaseous molecule to extrapolate it to the gas at room temperature - ArgusLab doesn't. While the heat of vaporization could be somewhat estimated, the heat of fusion isn't up to now (research topic) since it depends on every detail of molecule stacking in the solid, and the heat of formation of the solid is inaccessible to software.
So if you accept 40-50kcal uncertainty, several methods are possible. If not, only experimental values work - and these are too much neglected in the computer era.