Morse code is designed so that you can decode it with a binary tree

Morse code is ~180 years old, and was developed over a period of years. Figure that awareness of letter frequencies, Huffman encoding, etc. was pretty limited back then. And cruft accumulated in the developing standard. And a number of sub-optimal choices might have been made to address "very similar encoding of words with very different meanings" issues. And...

In real morse messages, you aren't always sending full words, most of the time you are sending prosigns (abbreviations) - Think like early SMS chat. "DE" means this is, you hand over with a "K", "SK" means you are going offline, "HIHI" is like LOL, "XYL" means significant other, etc...

Things like Q get used a lot in Morse for example (look up Q-Codes) compared to real English.

So, traditional letter frequencies are not applicable (Also because the code itself is language-agnostic)

Using a tree like this to decodee also sucks from a speed perspective. You need to learn the sounds /not/ count the dits/dahs if you want to get a good word per minute.

source: radio ham who knows morse and can do 20+ wpm.

A lot of weird stuff in morse and things also comes from older comms methods like Semaphore - The reason we have M in Scottish callsigns for ham radio is because M is the saltire (Scottish flag) symbol in Sempahore, for example.