your link budget parameters are fine
and the idea to use okumura hata is good too, that’s what you should do.
now the only problem is that you may want to use more margins:
1/ slow fading margin (the one you used is the “fast” fading margin), which is about 10dB. It takes into account a lognormal distribution of obstacles in the area, blocking the signal. It’s difficult to explain 🙂
2/ design margin, or safety margin, so that nobody will jump at your throat because your results were way too much optimistic, and there are coverage holes everywhere…
3/ body loss, which is in reality closer to 6 or 9dB than 3dB.
basically, you could remove 10+6+3
= 19dB to your original FSL result.
By the way, I didn’t check your computation, but you did both downlink and uplink ? then you took the worst FSL ?
also, don’t forget that if yo intend to use more than 2 TRX per cell, you should take into account combining losses. (about 4dB loss)