The memory card is able to respond within the first cycle. So according to several other board designs (and of course the famous newsletter) - dtack can be grounded.

I think you need to ensure the !WR pulse ends before asserting !DATACK. The classic way to do this is to have !DS control a 74LS164 shift register. See this circuit:

https://cdn.hackaday.io/files/295311263454304/J061_cct.png

There is a full explanation of how it works in the manual. You should be able to generate the timing you want very easily.

https://hackaday.io/project/29531-stebus-io-prototyping-board