[aerohead]

It's a 'success' because you've obtained valuable data you couldn't otherwise have obtained.
'Armchair-quarterbacking', I'd say that the first issue involves the 'cuff.' Let's say, it's a 12-inch radius. Imagine a 'clock face' 24-inches in diameter, laid directly over it.
If the flow over the roof and sides reaches the beginning of the 'cuff' at 12:00 o'clock, you're going to experience flow separation when the air reaches '4-seconds' past 12:00.
'Technically', this is where your attachment point would be.
Whatever angle chosen, it would begin at this location.
--------------------------------------------------------------------------------------
On simple, 'van-like' vehicles, with straight sides and roof, the lowest drag for 'straight angle', 'sharp-corner', boat-tailing has occurred with rather 'shallow' angles.
* The Ahmed body achieves minimum Cd at 9-degrees slope.
* The Howell et al. Windsor body 'sharp-cornered tail achieves an 18% drag reduction with 8-degrees top & sides, flat bottom.
* ATDynamics Trailer Tail uses 12.5-degrees top & sides, flat floor.
* Lawrence Livermore National Laboratory used 11-degrees, top & sides, with flat floor.
* I do not possess 'true-length' images for Clarkson University's tails, but they'd be worth the 'hunt.'
--------------------------------------------------------------------------------------
It seems logical, that changing from 14-degrees, off the 4-second location on the 'cuff', to perhaps 12.5-degrees, would get you in attached flow territory.