hello Tahar, 18S rRNA is not recommended for species-level discrimination. As for ITS, the cutoff is 97%. You can find more information in the paper I attached
Using ITS for species discrimination is a very unreliable and confusing methodology, I can say, it works better for genus level. If you already have a well known reference sequences for each individual species may be it works. But for better conclusion and acceptance with scientific community then better to use other specific primers for the give species.
First use ITS to separate genus level then once you get your interest of sequence collection then run them using specific primer that can separate the specie level.
Use a cutoff 95 to 98 depending on your sample size and goal of research.
Why don´t you try a combination of genes? I mean, ITS, 18S, 28S, b-tubulin, or others you can have access, in order to have a greater confidence of the fungal identity at species level.
ITS is recommended primary bar code for fungi, if someone get same ITS sequence for more than one species than go for ETF1 alpha. Please consult Stielow J, Lévesque C, Seifert K, Meyer W, Irinyi L, Smits D, Renfurm R, Verkley G, Groenewald M, Chaduli D 2015. Persoonia 35: 242−263. and best wishes to you