Here are my Anki stats as of this morning:

The 7062 seen cards in this deck contain:
1227 total unique kanji.
Jouyou: 1142 of 1945 (58.7%).
Jinmeiyou: 18 of 287 (6.3%).
67 non-jouyou kanji.

Jouyou levels:
Grade 1: 76 of 80 (95.0%).
Grade 2: 154 of 160 (96.2%).
Grade 3: 182 of 200 (91.0%).
Grade 4: 165 of 200 (82.5%).
Grade 5: 140 of 185 (75.7%).
Grade 6: 124 of 181 (68.5%).
JuniorHS: 301 of 939 (32.1%).

What does this lead me to think?
First, the 7062 seen cards represent only 3531 unique *facts*, since I have front and back.

So... first thought is that, even though each card has at least one, if not two, new vocabulary, I'm only adding a unique kanji every third fact, on average. That means most of my new words are now being made either with known kanji, or are hiragana/katakana words. To me this lends credence to using RTK lite as a beginning step, since so many of the everyday words are based on those first 1,000 or so kanji.

Notice, though, that I also have 67 non-jouyou kanji in my deck, while I still have more than 700 jouyou kanji not in anki yet. Hmmmm.... a step back here:

I have gathered most of my sentences from, at first, simple texts like Basic Japanese Sentence Patterns, All About Particles and the first bit of KO (though I got bored of that), and phrase books. Now I add from anime, movie, manga, go books and sentences from conversations ... "How do you say x?" or "Sorry, what does that mean?" I have no system except putting in sentences I find either useful or interesting. Yet by this somewhat random accumulation of words, almost 5 per cent of my kanji are non-jouyou, and I'm not even close to 10,000+ sentences. Since the number of jouyou is limited, this percentage can only increase, and to me shows the importance of learning non-jouyou kanji, even early on in your studies - especially the 50-100 that are so basic everyone wonders why they're not in the jouyou list.

And interestingly enough, though I have 1,227 unique kanji, I'm still missing four from Grade 1 level and six from Grade 2. At first this suggests that random gathering of sentences leaves holes in your learning. Sometimes people suggest on forums that you might miss out on some simple words if you're not systematic. Yet I've seen the list of Grade 1-2 kanji, and I know them all - I guess there's a few I've never seen in a sentence interesting enough to keep. On the other hand, just by randomly collecting sentences, eventually all the basic ones will get included, in one sentence or another. Random acquisition seems to work as effectively in the long run as systematic accumulation.

Just some thoughts...

I think it's best to do both. ;p

Kanji statistics
The 2917 seen cards in this deck contain:
1863 total unique kanji.
Jouyou: 1676 of 1945 (86.2%).
Jinmeiyou: 41 of 287 (14.3%).
146 non-jouyou kanji.
Jouyou levels:
Grade 1: 80 of 80 (100.0%).
Grade 2: 160 of 160 (100.0%).
Grade 3: 200 of 200 (100.0%).
Grade 4: 198 of 200 (99.0%).
Grade 5: 185 of 185 (100.0%).
Grade 6: 180 of 181 (99.4%).
JuniorHS: 673 of 939 (71.7%).

These are my stats.

I'm posting them because I have a similar number of sentences: 2917, just ~600 less than you. But I'm already with 1863 seen kanji.

A lot of my sentences come from dictionaries, but the vocabulary which I search to find these sentences come 99% from authentic texts.
I'd say that ~50 sentences are "artificial", in a sense that I used the missing kanji from anki to google sentences for these kanji. So these ~50+ kanji that are in my deck may be not from everyday usage.

So maybe, kanji light might be enough, but maybe a kanji lite+, with a little more kanji than kanji lite would be better (around ~1500?).

JRTK is always a good option. I like the deck for many reasons. For one thing, it's all production, and it focuses on each kanji in RTK*. Therefore, you can, like I have, trash a lot of the old production cards in your deck that may be unnecessary.

Another good thing is that you can always go in and change either the sentence or "keyword" for each kanji depending on your needs. I often replace the default sentences with more friendly ones from my readings, and I also change some of the default keywords as well. I've also found that it keeps my mind away from kanji statistics, in that I know that I'll have the full list done by the time I get to the end, plus production for each kanji.

I still do the bulk of my sentence mining from whatever interests me, (right now, university physics), I just simply add cards from JRTK little by little as I study. I've also eliminated the need to study RTK English keywords. Also, sometimes it's just nice to have cards that you can drop in to bulk up your deck, without having to type them up.

*I haven't checked the updated JRTK, but the one I downloaded three months ago is missing about 250 - 300 kanji, although I use the model to fill in some of the gaps as I go.

-Edit-

I checked the spreadsheet, and the JRTK deck currently has 1871/2042 kanji available, missing 171 kanji from RTK 1

What is JRTK? Japanese keywords?
I guess that would be specially good for people that already know a lot of Japanese and are getting back to pick up the Kanji left behind.

JRTK info found here:

http://wrightak.googlepages.com/afterrtk12

@ mentat_kgs
Interesting stats to compare. You're much more kanji heavy than me, it appears. I actually have a lot of katakana words (boo, hiss), so that makes sense.

I usually just maintain my RTK knowledge by studying sentences, it's just so easy to rememorize any that I slightly forget by the time I come across a word with that kanji. Haven't opened my RTK deck in a month or so.