This is actually quite a big subject, so I'll go through at a high level at first.
First, as you pointed out, it's a nice system with a ton of great JAPANESE games. But unfortunately, its marketing in North America (including the library chosen for publication here) was overall disappointing, so that even today, it's considered a niche system.
Next, you will see (not just on this site) that in North America, NES and SNES - due to their popularity - drive the largest amount of discussion, which in turn leads to the largest developer community. And game players. And people interested in doing translations.
You can see the disparity by searching the number of translation for these systems - Nintendo dwarfs TG-16 (or PC Engine, as I will call it from here on).
In order to have get a translation done, usually there is a technical person and a translator. And both of them need to have a lot of perseverance. There are several strong technical people in the PC Engine community, but we suffer from a dearth of translators. As you mentioned in your post, the Startling Odyssey II translation was completed without a translator. (As an aside, this fact generated a storm of nasty comments on this board about whether it could even be considered a translation - which is why you won't find it here).
Now, as for the differences between HuCard and CDROM, the first and most important to consider is SIZE. A CDROM game generally has a much larger script to translate, which leads to commitment issues on the sides of both the translator and the technical person.
Of course, there are technical differences as well, but basically if you don't start a project, you won't need to worry about those. I'll give you a high-level on those next.
First, the PC Engine CDROMs are not stored in an ISO-9660 format as some modern games are. Their disc format predates even High Sierra, so NEC chose to just stick a large blob of data there. Consider it a giant ROM, which is loaded in sections, so you still need to observe the same practises you would on a ROM - except you can't arbitrarily extend the ROM and be able to map the new area. This is because a new ROM area can be mapped instantly (nanoseconds), but an extension of the data track on a PC Engine disc needs to consider seek time, which will basically be hundreds of milliseconds.
On the bright side, many CDROM games were written thinking, "I have so much space, I don't need compression", so the text can often be found easily (it is often stored in SJIS, because the CDROM system card contained fonts for this). The double-byte encoding works to your advantage, as long as you switch to ASCII and alter the print routine to match.
I have written a little 'starter series' on PC Engine translation here:http://pcengine.proboards.com/thread/557/translation-patch-post-level-overviewhttp://pcengine.proboards.com/thread/561/translation-patch-isolating-print-functionhttp://pcengine.proboards.com/thread/583/translation-patch-print-function-modificationhttp://pcengine.proboards.com/thread/595/translation-patch-post-extracting-script
...and as a follow-up, I even made the print function modification easier with this hack:http://pcengine.proboards.com/thread/632/translations-little-bit-easier
Hopefully, this helps explain a lot of the things you asked about. Let me know if there are more questions on the subject.