What you are seeing does make sense. My clone only decodes up to A9 which allows for address up to 0x3FF which I believe is the official upper bound for expansion card IO addresses for IBM 5150 and 5160 PCs at least.
Yes, I've just looked through the manual of the 5150 here:
http://minuszerodegrees.net/manuals/IBM_5150_Technical_Reference_6025005_AUG81.pdf
Eg on the Game Adapter part, they say that only A0-A9 are used (page 2-119).
They clarify that the card responds to IO addresses x'201', so that would be exactly what the Game Blaster clone does.
If you look eg at the Asynchronous Communications Adapter, the table on page 2-125 also only lists A0-A9.
A bit strange then, that IBM/Yamaha chose to use the address range of 2A20/2A30 for the Music Feature Card. It came out in 1987, and was at least meant to be used in 5160s, possibly 5150s as well.
I don't think I have ever encountered a card that went beyond 0x399 before. It would be interesting if someone with an original Game Blaster or CMS could look at their card and see if the address lines above A9 (A10 through A19) are connected to anything. I would be surprised if they are but who knows.
I agree. In fact, I have a Sound Blaster Pro 2.0. It would not surprise me if this card also clashed with the IBM Music Feature Card if I configure it to 220h. I will test that out sometime.
By the way, what I said above:
Then I reconfigured the Game Blaster to a base address of 210h, and put it back in. Writing and reading FF was no problem... however, the second test is to write and read 0.
Doing this yielded 'E5' when reading back.
I think this was my mistake. My detection routine did not reset the card first, so reading back the register was returning an error code, I believe.
I added a reset to my routine, and now it detects the card properly (writing and reading FF first, then writing and reading 0). I've tried it with the Game Blaster jumper in every position, from 210h to 260h, and only at 220h it didn't detect it.