The most common standards are -20 and -18, there are some people who use -22 though. It's an issue of how much head room you want while you're mixing. You can mix to whatever reference you like, just know that it will affect your mix decisions. For example, if you mix to a -22 reference your mix will likely be quieter than if you mixed it at a -18 reference.

Setting your digital/electrical reference level to match a specific SPL is also partially dictated by the volume (meaning size...you know, length X width x height) of the room you're mixing in:

Greater than 20,000 cubic feet (566 cubic meters) - 85dB-SPL
10,000 to 19,000 cubic feet (283 to 565 cubic meters) - 82dB-SPL
5,000 to 9,999 cubic feet (142 to 282 cubic meters) - 80dB-SPL
1,500 to 4,999 cubic feet (42 to 141 cubic meters) - 78dB-SPL
Less than 1,499 cubic feet (41 or less cubic meters) - 76dB-SPL

In a normal theatrical environment/stage it should be either -18dBFS = 85dBSPL or -20dBFS = 85dBSPL (I believe -20 is the commonly-accepted standard now). For TV its the same but to 79dBSPL. I'm curious where the 32 comes from. That's a very oddball number unless There's something I'm missing entirely.

-20 dB Full Scale is the commonly accepted standard.

I believe I located the paper you were referring to (in Italian):

http://www.atic-ntc.org/Notizie/Gestione%20del%20Suono%20nella%20Produzione%20Audio%20per%20la%20Cinematografia.pdf

Section in question:

Osservazione: Dopo aver studiato il problema e misurato rumori reali in ambienti dal vero, e quindi indicativi delle esigenze di dinamica di un film, si è rilevato che è consigliabile fissare una corrispondenza tra un rumore rosa a 85 dB SPL (pesato C) nel punto di ripresa, limitato in banda (20 Hz, 20 KHz), e un livello RMS digitale equivalente a quello fatto segnare da una sinusoide a -32 dB FS di picco. I parametri di allineamento qui riportati sono da considerarsi validi non solo per la presa diretta ma anche per la registrazione di musica, effetti e doppiaggio. Sarà poi in fase di mixage che tale dinamica, realistica, sarà adeguata agli standard di consegna del prodotto. Tale indicazione di massima può ritenersi un utile punto di riferimento, che favorirà ampia dinamica e lavorazioni di postproduzione omogenee e agevoli dal montaggio al mix. L'audio cinematografico non e’ neppure in grado di riprodurre realisticamente picchi quali quelli prodotti da un semplice battimano, che presenta appunto picchi di 30 dB superiori a quelli del dialogo, richiedendo almeno 20 dB di headroom in piu’.

I was able to get a rough online translation:

...it was noted that it was advisable to attach a correspondence between a pink noise at 85 dB SPL (weighed (C) at the point of resumption, limited in bandwidth (20 Hz, 20 KHz), and a RMS level digital equivalent to that reported by a sinusoid to -32dB FS peak...

I also have never heard of -32 dBFS as a reference level - in the US we almost exclusively use -20 and EU almost exclusively -18. I woud be interested to understand more from this paper and see what the author's basis is. If someone could translate that would be great. Thanks.