with which button do you jump? I downloaded it now but can't figure out

Oh, the thing in bold, that makes sense then >.<

BTW, i love the greyscale

Awesome, I will so download it!

Cool, what do you plan to use the font for?

Man, I really going to have to put it back on my calc(Sorry, I haven't had really the time yet to do so).

EDIT:Done  (With done I don't mean I finished the game, I just mean I put it on my calc)

I just completed the last level of Jumpman Junior. It is called Walls:



The walls can push you around, but they are not impenetrable. You can walk or jump against the wall to push through it. 

* the standard fonts of the TI-68k/AMS platform look more familiar to users than other fonts;
* the levels' names would clearly to be easier to read under F_6x8, F_8x10 or Apogee;
* as you're using program-specific data and your own drawing routine for Apogee, you might as well use it a bit more widely than just a single level - that way, less space will feel wasted

If I recall correctly, the standard text drawing routines are extremely slow.

Even on AMS 1.xx, but it's indeed much worse on AMS 2.xx and 3.xx, due to a stupid implementation for the very rarely used ability to redefine fonts. I "fixed" it in tiosmod+amspatch by hard-coding the fonts, which makes AMS 2.xx & 3.xx character drawing performance at least go back to approximately AMS 1.xx performance levels.
Third parties can draw characters 3-4 times faster than AMS 1.xx / patched AMS 2.xx & 3.xx can.

But I do recall someone discovered a way to access the data of the TI fonts...

That was me, indeed, as the final improvement to a method originally devised by Tom, which required a fairly slow setup phase of drawing characters into a temporary buffer
Most TICT programs which need to draw strings quickly have been converted to that method. The TICT eBook reader is probably the canonical implementation; that said, it does not support all font sizes, while some others do, so I'd have to check.
Fast integer to string conversion and fast font drawing routines are among the refused contributions to TIGCC, notwithstanding their clear use cases. Users never got around to benefit from them in the standard development environment, and I'm not spending time integrating them into GCC4TI either: the number of users is way too low, and the remaining ones who are interested in alternate string drawing, like you, can cope with it by themselves

but are they in an expected "sprite" data format to be easily drawn using the Extgraph libraries for speed?

Third-party character / string drawing routines are just fancy sprite routines (with a bit of extra work for F_4x6, as you suspected). Usually, I didn't add a dependency on ExtGraph's sprite routines: the character sprite drawing was inlined, to save the cost of a function call and return.

I agree that TI's quasi 4x6 font looks better, but it's width varies per character.

Yup.

It seems there has to be some sort of table containing width information for every single character (possibly embedded in the font data itself?) -- ugh!

Exactly, embedded in the font data itself

I only drew 64 of the 256 non-masked characters though -- to save space (512 bytes total).

Yup.

I could trim this down even further if necessary.

Well, it wouldn't save that much, so I'm not convinced you should bother until you're really starved for space