I think that you are trying to read too much into things and, as a consequence, making it harder to understand than it is ...
when i reduce level segment by 0.8%
you don't have to worry about any reduction - it's built into the AFM calculations.
So,
(a) you climb (with real world OEI performance) in the second segment to, say, somewhere around 400 ft. However, the AFM takes you (at net performance) to a lower height something less than 300 ft. This lower height is relevant to the obstacles, not your altimeter height.
(b) then, at the gross altimeter height you accelerate, nominally level, clean up and eventually achieve final climb speed. However, the NFP acceleration is done (for obstacles) at the lower NFP height and, being a degraded acceleration, takes a longer distance. End result is that, at the end of the NFP third segment, the real aircraft is well on the way into the fourth segment. When you end up at 1500 ft in the aircraft, the NFP is still way, way underneath you and the NFP doesn't achieve 1500 ft until a considerable distance further on .. and you base obstacle clearance on this lower profile.
(c) for the pilot not wanting to die on the surface of the obstacle, the main place to worry, and fly really accurately, is in the early second segment .. the further away you get from the runway, the higher the aircraft gets above the obstacles.
(d) a lot of folk have this idea about the aircraft OEI being 35 ft above obstacles miles out from the runway. Not so - the NFP might be in that predicament down there ... you are progressively and considerably above.
I go below the level segment height to a certain point.
No.
You will be flying level or, in the higher performing aircraft, a shallow climb. There is no intent that either the aircraft or the NFP calculations see a descent.
However from the pilot point of view certification is secondary whereas compliance with the given instrument procedures is primary target.
The certification (Design Standards) aren't overly interested in the missed approach - that being an operational concern. Caveat relates to the landing and missed approach WAT requirements but that's about the extent of it. Once you observe the WAT limits in determining the TOW, you have addressed the certification bits.
PANS OPS looks at how one might go about designing procedures and such like.
The silly situation is to let the aircraft head off down into the letdown WITHOUT having had a looksee at what you might be able to do IF one quits (or has already quit). The AFM doesn't give you much in the way of an easy way to calculate a missed approach climb but, with a bit of engineering pushing and pulling, the available information can be reworked to give you something equivalent to a takeoff analysis. Might even need a bit of simple flight testing to get some quantitative data to fill in the gaps.
meaning that the required missed approach gradient must be achieved till the final phase.
.. or you do something more rational which suits the aircraft and keeps you above the rocky bits.
In practical terms I have no idea when the min of obstacle clearance of 164 ft is achieved except when I'm at the MSA or final level off ALT, whichever is lower.
Which is the problem with the operational side of things - unless you do a rational analysis you are playing Russian Roulette, pure and simple.
The question is how was the GA performance calculated to follow published SID SE?
.. it wasn't .. that's a problem belonga the pilot and operator. Feasibly, the operator OUGHT to make sure that any critical approaches have been run through a rational ops eng analysis to come up with a specific procedure to get you out of there if it all goes pear shaped ...
I suppose since the company gives me the choice it's all been taken care of
.. now, that's a BIG assumption on your part ... I would incline to the view that the company which prescribes step A, step B, step C ... puts the pilot in the better situation .. presuming that those steps are based on a rational engineering analysis.