OpenType features with chars from different Unicode blocks

[UCS_Font_man]

Hi everyone.
Here's my problem, which, I do much hope, you will help me to solve.
I work with Old Cyrillic text and use specially designed Font (UCS) to represent it in a right way. Its main feature, that it uses Basic Latin Unicode Block codepoints (and numbers also) to store additional Cyrillic characters. It means, that in this Font - Latin letters are no more, but they are replaced by those special Cyrillic characters. So, if ordinary font is applied to such text, words in it represents consisting of Latin/Cyrillic/Numeric characters jointly, here's the sample of text:
"Во дни6 же w4ны пріи1де їwaннъ кrти1тель, проповёдаz въ пустhни їудeйстэй и3 глаг0лz: покaйтесz, прибли1жибосz цrтвіе нбcное. Сeй бо є4сть рэчeнный и3сaіемъ прbр0комъ, глаг0лющимъ: глaсъ вопію1щагw въ пустhни: ўгот0вайте пyть гDень, пр†вы твори1те стєзи2 є3гw2."
But its not the problem, - we use it this way for long time already, and its all fine. My problem began, when I tried to add some OpenType features to this UCS font: I've implemented (GPOS\GSUB) lookups and features for it (I've read tutorials and threads on this forum and it helped me great, thanks to all) and it worked, but kind of partly. I mean, it works only, if characters, which are engaged, are from the same Unicode plane, like Latin with Latin, and Cyrillic with Cyrillic (as I said, I tried all kind of features with positioning and substitutions and it worked well with the such chars.); but not in mixture, like Lat/Cyr or Cyr/Lat.
For example: Lat. "A" with Lat. "B" - works, but Lat."A" with Cyr. "Б" - doesn't.; Cyr."б" with Cyr."и"- works, but Cyr. "б" with Lat. "с" - doesn't. I hope, I've explained the problem clearly, sorry for my poor English. I will be very thankful for your suggestions / solutions.

[Erwin Denissen]

That is just how OpenType layout features work. They are bound to a script.

I strongly recommend to move forward and correct the codeponts to the actual Unicode values.

[UCS_Font_man]

So, there is no way to overcome this bounds?
Maybe some trick