All depends on how strong you want the archetyping force to be. . .with a spread of costs from 1-20 (or higher) it's very strong, with a spread of costs like 1-4 it's very weak. . .1-11 would fall somewhere in the middle. I tend to moderate that by what professions I allow.
Diminishing returns on ranks tends to push people to spread out and become more round. (Semis, by being already spread, tend to take longer to reach diminishing returns)
I suspect that if one created a series of 50th level characters, you'd see how a 1-4 spread would result in very round characters that can do a lot of things very well, and everything at least moderately well. . .while the 1-20 spread would still see archetype focus having a strong effect. . .
There's really no right or wrong answer there, it's more a matter of how strong you want archetyping to be. I've played RM with no-semis allowed, which makes for rigid archetyping, or with "All characters are no professions" which leads to weak archetypes. (Though you'd be surprised actually, how much archetyping players enforce on themselves, especially at lower levels, in the "All NP environment.")
As I see it, there are 3 factors in play for archetyping.
The cost spreads (Much discussed already)
The Size of the Skill Set (Fewer skills leads to more round characters, more skills to stronger archetypes)
Diminishing returns (The prod to get out of archetyping and become more round)
Semis, as is, with a relatively small skill set, tend toward round, slow development, but I don't necissarily see that as a problem, in exchange for being slow developing, they have more abilities and options.
Much of the complaints to both sides on semis don't end up directly related to cost, but to cost-benefit. . .Most of the core semi spell lists are weaker versions of pure lists, so even with the 2.2 option for individual development in place, they're paying twice as much DP for a much weaker spell, so even if the cost is x2, the cost-benefit is more like x3 or x5. (Compare the Ranger lists to the Animist lists for instance). . .then many of the semis that people complained were too strong was not due to direct costs, but because their lists were stronger, as good or better than the core pures. (Compare "Guardian's Ways" to the rest of the ranger lists for instance.). . .with a game like RM, it's often hard to define the cost-benefit broadly.
That kind of makes the semis a mixed bag. . .in many rules set ups you might be better off playing an Animist with a lot of combat skills rather than a Ranger, but probably you're better off playing a Venturer than trying to make a martial skill focused Mentalist.