Unclear

REACHED the following rather unclear position today against a junior in round five of the Hastings Masters.

Black has just played an "only" move, 18...Qf7-d7, in Spanton (1954) - Thivyaa Rahulan (1814)

Note that the aggressive 18...Qh5? would have run into 19.h4, when Black would have no good way of saving her c pawn, eg 19...c6? 20.Bf3 etc.
After the move played in the game, 18...Qd7, I probably should have chosen the solid 19.f4.
Instead, I went for the risky 19.Bh3!?
I felt Black had three main tries.
Firstly, the cautious 19...Kh8 really must be answered by 20.f4, when my main analysis engines, Stockfish10 and Komodo9, reckon White has the upper hand.
The move that worried me was Black sacing the exchange with 19...Rxh3!? After the forced 20.Nxh3, I had concentrated my analysis on 20...f4, when I was probably going to play 21.Ng5 (although I was also seriously considering a counter sac with 21.Nhxf4!?). I felt that if the worst came to the worst, I would be able to give back the exchange, but it certainly looked hairy, and the engines are divided - Stockfish10 believes Black has full compensation for the exchange while Komodo9 gives White the upper hand.
However, another continuation after 20.Nxh3 is 20...c6, which I did not give enough attention to. After 21.Nc3, the engines at various points pick 21...h6, 21...f4 and 21...d5, with verdicts varying from equal through White being slightly better to White having the upper hand.
In any event, it is safe to say 19...Rxh3!? is complicated.
The game saw 19...c6? but TR had missed the reply 20.Bxf5! (1-0, 49 moves)