I'm sure I don't have a definitive answer to offer, but I'll share what I've learned being in a similar position -- self-taught pianist and composer for about 14 years. I haven't read as much as you, but I've probably improvised more.
One of the most valuable things has been alternation between book learning and finger learning. My method for the longest time was to just play and play and try new things and see what I liked and investigate anything that was interesting. Often, songs would develop through improperly or accidentally struck keys! For example, the bridge in one song started from meaning to hit D, A/C#, F#m/C#, which was part of the standard melody line, but for that last chord played C#sus4 instead, transitioned into F#m/C#, and from there back to D. Sounded nice so I developed it further!
But now and then I will take the time to learn theory. Sometimes experimentation leads to original composition, but very often it just chips away (slowly!) at well-established and basic principles. Since it's possible to digest a principle in an hour with a good chapter instead of over weeks of experimentation, I now try to get these "infusions" of knowledge and start playing on a structure. Similarly, I will play other people's songs, sometimes spending a day or a week working through new songs or sheet music, or more if I'm finding it helpful, and notice what I like best in it and intentionally improvise on the base that they've given (whether it's their song structure, their rhythmic pattern, their chord progression, their melody line, their harmony style). This provides fresh "mutations" for my basic compositional DNA.
In the same way, many of my new songs actually start the same as an existing one; they'll have the same intro, for example (say, 4 or 8 bars of a rhythm or a chord progression). This provides a runway for the creative process. Then I just go in a different direction, again experimenting, whether it's physically or mentally: Physically, shall I try this higher octave to see what it has to offer? Mentally, shall I try a new minor chord where my instincts tell me to use the familiar major one? Physically, shall I do a chromatic trill on this note to open up the scale? Mentally, shall I modify the harmony to work in snatches of counterpoint?
Now, sometimes I do work intentionally instead of experimentally. I sit at my score without my hands on the keyboard and think, "What is needed for this climax is a modulation up a major third. I'll need to use two measures to transition that, which I can give to the flute and accent the key 'pivot' notes with the horn," or something. But because I'm not as well-read, I do more discovering the other way. You might find that your scale is balanced more to this sort of thing.
Also, I record a lot of content to make sure I don't lose ideas, but I do at least 3 or 4 times as much playing without recording. This playing is other people's music or favourite songs of mine I know aren't quite developed. Like a toddler babbling, this is the exploration, the dwelling and steeping in the piece until something new emerges. As the cliché goes, the voice of inspiration comes when you least expect it, but I find it rarely comes if I'm not listening. How I "listen" is often just to spend that patient time with a piece.
Happy composing!
Appendix on improvising
In reply to your comment: I'm not sure the process is easily transferable, and I wouldn't say that I improve "in order to learn something new". What I meant above is that I learn either by intentionally reading or by accidentally playing, usually after getting to the point of boredom with whatever framework I started with. Instead, I improvise to get ideas for new pieces. :)
That said, for what it's worth, here's the sort of thing I do.
Pick a key — whether this is done by fingers finding familiar keys or intentionally choosing one doesn't matter.
Start with a few chords or a more or less random melody line.
Whichever was chosen first, add the other.
Develop! Repeat one element while testing the other. (One thing I really enjoyed about Miles Davis' ensembles was how they would sometimes have a constant rhythm while changing the melody, and sometimes change the rhythm while having a constant melody!)
Make your way gradually towards a melody that you would call the central idea, and a consistent chord progression for it. Play around in the space between. New chords, new melody fragments within the key. Make your way back to that central idea every now and then, taking longer ways round till you return to it, which immediately lends it new thematic interest as the context around it changes. Try to build an arc so that you remain emotionally interested.
This process is generally enough to get an idea that I can then begin to more intentionally elaborate.
Here's a snippet of the opening of an old improvisation of mine (nowadays I hope my timing is better... my recording quality certainly is!). I don't know if it'll be your cup of tea, and it may not actually be representative of the above since it's 3 minutes of what became 26 and doesn't even begin to touch on the main theme, but for me it represents a good balance between a repeating structural element and playing minor variations to get somewhere new and find something I like. Towards the end of this clip you can hear where I found it got samey and decided to throw in a non-diatonic change in the hopes of slightly changing the game.
2015 improv bite
