Although dragging this out seems fairly pointless (since those who can see the difficulty, can see that it isn't relevant anymore, now that we've got a bigger range of values to work with, and those who can't see the difficulty will be able to change the values they pick for bonuses when balancing things so that there doesn't end up a problem, so they'll never actually need to see the difficulty explicitly, so all told this is a non-issue) , for cussedness' sake I'll have one last go.
[note about terms of discussion]
There are many factors to be considered here. Empire A produces A(t) at turn t, empire B B(t). There are the relative (a/b) and absolute (A-B)advantages that an empire might have over another at a particular instant, there is the sum of that advantage over time, SUM[t](A(t)/B(t)). and then there is the cumulative advantage an empire might be able to gain after time from starting out in front I.E A(t2) c.f. B(t2).
It's important not to mix these up, overwise you end up thinking you've countered an argument, when all you've done is made an argument about something totally different.
My "there are at most six fully stackable techs with equally desireable bonuses" statement was meant to convey just that. i.e not that we should outlaw techs that give bonuses of +7, or that they would be unusable. That's silly. It just meant that if you had +16, and your last bonus was +8, then you'd be looking for a bonus of about +16 for a similar relative cost, to have the same magnitude of effect, anything with a lower bonus would, other things being equal - which they rarely are, be less significant in your considerations.
Now, given that we now have a reasonable range of numbers to work with, we have several reasonable choices. Either
a) keep the costs of each additional +1 bonus roughly constant in absolute terms, so that in relative terms getting the extra bonuses becomes less costly, so the cost for an equivalent relative bonus is pretty constant in relative terms. In other words make player fight like hell for piddly +1 and +2's early game, but give away +5's and +6's like confetti in late game...
b) increase the mapping function from bonuses to effects roughly geometrically, and keep the cost to get each +1 (or ~+10 given the new range) roughly equivalent in relative terms (i.e. later techs give similar scales of bonuses to early techs, but they cost much more, because they generally stack with the early game ones you are expected to have got, and so get you to ever more dizzy heights of total bonuses).
or, the route we will probably take if we don't explicitly consider the issue
c) a bit of both, late game techs cost a bit more, give a bit better bonuses, the two together make most techs fall on a geometric absolute cost/benefit path (i.e. a constant relative cost/benefit path), and we play with both numbers constantly until they get somewhere near balanced.
Like I said, it really doesn't matter, since the first balancing pass of method c) we probably be to get us back to the effect of a) or b) anyway, from which point on they are all pretty much the same in practical terms, if not conceptually.
p.s. For anyone who really thinks that it's the absolute differences that matter, not the relative ones, imagine what the variety of reponses of people who earn different amounts of money would be to a suggestion that all income or expenditure related taxes should be scrapped and replaced by a fixed tax of $2000 per year... (n.b. yes, this is a realism argument, and as such a bad thing, however the underlying economics principle of 'relative marginal value' applies to the game as well, and the real world example just highlights it.)