I'm not sure where the 71% comes from, but I will examine this direction- because you're right, I didn't really account for strong characters using weaker maho, though part of the draw is also supposed to be weaker characters using stronger maho- a quick path to power, as it were.
So, I'll use a specific example, using the (1d5-3+2xRank) taint gain idea I put forward- I have a School Rank 4 character with highest ring 4. I cast a rank 2 spell, and I spend a void point to soak the taint. I'm now rolling 5k5 to soak 1d5+1 taint- so TN30 (at a 49% chance of success) to be guaranteed to be taint free- I won't take those odds, or even the 71% chance to soak 5 points, so I'll aim to soak four. That still means that 40% of the time I walk away with a point or two of taint- and there's still a 10% chance of failing to soak four, which means that I'm likely to eat the whole 1d5+1 at some point, if I use Maho too often.
Same example, using your 1d5+2xRank- that same character is now rolling 5k5 to soak 1d5+3 taint. If I want to be guaranteed to soak it all, I have to roll TN 40- a 17% chance of that, so no. Even if I soak 4 points of taint as in the previous example, I'm leaving with an average of 2 points of taint every time. That's a rank 4 character suffering a higher-than-book-standard rate of taint gain from a spell two ranks lower.
If I have a specialist character with Ring 5 at rank 4, then yeah, the odds improve- but not a great deal. The ~90% threshhold that I'd want to aim for now soaks 5 points of taint instead of 4, which mean that I'll occasionally get a point of taint from that rank 2 spell in my setup (except for that 10% of the time that I fail the soak roll), or I'll usually get 1-3 points of taint from a rank 2 spell, except for that ten percent of the time that the taint of Jigoku invades my body.
Now let's look at a rank 1 character casting a rank 5 maho spell. This is a thing that can happen, which is (again) part of why Maho is supposed to be seductive. Let's say it's a favoured ring at rank 3, so rolling 4k4 (with Void) to soak either 1d5+7 or 1d5+10. Core would have him take four points of taint and call it good. With my system, he walks away with 5-9 points- that's a lot- and with your system he walks away with 8-12 points. An average of an entire taint rank
at a single go.
I understand not wanting to make Maho safe- even the weaker Maho for high powered characters. This is probably why the game just throws out flat values.
So, here's a different idea (just brainstorming here)- use the soak mechanic you have, each Maho spell causes Taint at a rate of (Maho rank) rolled d5s, keeping one die. 5's explode. This means that a rank 1 Maho spell has an 80% chance of being 4 or less points of taint (not too hard to soak), but a 20% chance of being 6-11+ or more points of taint (equal odds 6, 7,8,9,11+, which values are much harder to soak). A rank 5 Maho spell will probably
be seven points (unsoaked) or less- but there's a 18.5% chance of it being 11 points or more (interestingly, the odds here are weighted slightly towards the higher
values, with a 3.9% chance of seeing a 16+, descending to a 3.5% chance of seeing 11). This, to me, exemplifies the seductive nature of Maho- no matter what rank Maho, you're only soaking one (exploding) d5. The odds of the Taint being manageable are really good... except for when you roll those four 5's in a row. Then you're screwed. Low risk, but very high stakes.
I used the Troll dice roller
to examine the odds, along with the Lynks L5R probability site.
If you want to examine the (Rank)k1 in d5's method of Taint calcs, go to the Troll site and copy the following into the dialog box- then click on "calculate probabilities." That will show you the calcs for 5k1 in d5s- if you want to change the number of unkept dice, just change the 5 next to the # sign to however many dice you want to do- I looked at 1 die, 2 dice and 5 dice, mostly.
\ Taint roll
sum (largest 1 5#(sum accumulate x := d5 while x=5))