Currently, because spell cost scales with maximum magic, but the regeneration rate doesn't, you are actively punished for investing additional sugar lumps into magic towers, or building more towers than the minimum required to cast the spell you want.
Sure you get more burst potential (some spells can be cast up to 4 times off a full tank!) But it overall takes longer to regenerate the magic for a single cast. Mostly, there are a bunch of weird optimization points that are hard to predict, but they cluster around having small maximum magic charge.
Off the top of my head, there are a couple of different options I can see to fix this:
1) Scale the charge rate with maximum magic: Just multiply the current charge rate by [current maximum magic] / 5. Now it never takes longer to regenerate a spell's casting cost than it did when you had the minimum magic required to cast it.
Upsides: Quick and easy, fixes the problem.
Downsides: no upper bound on magic recharge rate, technically possible to be able to cast spells arbitrarily quickly.
2) Option (1), but also rejigger the whole recharge system so that the charge rate is dependent on the current percentage full the magic meter is -- instead of 0.01 magic/sec being at the fixed point of (I assume) 0.5 magic, and 0.02 magic/sec being somewhere between 6 and 7, these points would move based on what your max magic level is.
Upsides: Fixes the problem without introducing any new balancing issues.
Downsides: lots of math and decisions about where each tick-mark should be on the meter. Still no reason to upgrade past the ability to cast the most expensive spell. Spell system now feels boring and samey, because the numbers don't mean anything.
3) Make spell costs fixed. Rebalance the whole system around fixed costs instead of scaling.
Upsides: Much easier to balance around.
Downsides: Complete redesign throws out all existing work.
4) Invert the spell charging rate. Make magic recharge faster the closer it is to empty, and slow to a crawl as it nears the maximum. Like in (2), the charge rate has to be based on how full the bar is, rather than fixed numbers.
Upsides: Fixes the problem. Cheap spells are much faster to recast than expensive spells.
Downsides: Radically changes balance. Again, some math work and design decisions to be made around scaling the recharge rate to the fullness of the meter. Depending on precise decisions, may still be no reason to invest in increasing max magic past a certain point.
5) Multiply the magic recharge rate by the upgrade level of magic towers.
Upsides: Easy. There's now a reason to invest more sugar lumps after unlocking the minigame.
Downsides: I'm not really seeing any? (For all I know you did this already and the wiki just hasn't updated with the information.)
6) Something with D&D-style spell slots, maybe?
Upsides: More nerdy
Downsides: I haven't thought this one through in the slightest.