In another post about the de-facto standard use of Ecclesiastical Latin pronunciation in singing, I included a postscript querying whether excelsis should be pronounced [ɛksʧɛlsis] or [ɛkʃɛlsis]. There seems to some doubt about the matter; what I've typically heard is the former, but the latter seems to be supported by at least some references. Is there anything that can claim to give an authoritative answer, or is it simply a matter of taste?
In Rules for interpretation > IX. The reading and pronunciation of liturgical Latin > Consonants (page xxxvii / 37 in the pdf linked) it reads:
SC before the same vowels [a, ae, oe, i, y] is pronounced like Sh in shed,
which we already knew. Then, the interesting part is on the next page:
XC before e, ae, oe, i, y - KSH.
e.g. Excelsis = ek-shel-sees.
Before others vowels XC has the ordinary hard sound of the letters composing it
e.g. excussórum = eks-coos-so-room.
I'm not aware of a more authoritative source than the Liber usualis regarding Ecclesiastical Latin (though there could be, e.g. in the Missal or some decree/motu proprio). BUT, there could be more details in the Latin version of the rubrics, which I haven't been able to find yet.
Update/Note: I plan to ask someone more knowledgeable than me (perhaps even a 70-something priest) about the authority level of the Liber usualis. But currently I think that even if it was not officially lex orandi (I'm not sure about that), it was very widespread in the decades prior to the Second Vatican Council and the liturgical reform (the linked edition is from 1961). At that time Ecclesiastical pronunciation was already in effect, and unity and authority were taken in a seriously formal way in the Church. Personally, I doubt the editors would have been willing to add their own personal taste to the detriment of unity, but I could be mistaken.
Bonus track: the rule for TI literally reads:
TI standing before a vowel and following any letter (except S, X, T) is pronounced tsee.
Otherwise the T is like the English T.
Therefore, commixtio > [com'miks.ti.o]! Anyway, the wording makes me very uneasy because English [t] is aspirated and I've always assumed Latin [t] is not. I just hope the distinction was omitted for the sake of simplicity.
From the other question, it seems you agree that starting from the end of the 19th century, after the recommendations of Pope Pius X, the Italian pronunciation became the standard one for E. Latin. If you do, then the way excelsior (a loan word) is pronounced in Italian answers your question - [ksʧ], and indeed Wiktionary abides by such standard for the Ecclesiastical pronunciation of excelsus. If instead you don't make this assumption, then it depends on taste and geography.
UPDATE: In agreement with @sumelic's contribution, I've finally got an article that confirms the recommended pronunciation to be the Italian one, i.e. [ksʧ]:
Pronunciación de los textos latinos puestos en música. Estudio práctico para la interpretación de la música española (Pronunciation of Latin texts put to music. Practical study for the interpretation of Spanish music) published by Francisco Javier Estrada Ramiro on Nassarre : Revista aragonésa de musicología, ISSN 0213-7305, Vol. 24, Nº 1, 2008, pages 59-96.
It is available here, in Spanish. On page 77 one may read
Algunos han postulado dos pronunciaciones en el latín a la italiana: la clásica del siglo I, hasta el s. III d. C., y la italiana a partir de dicho siglo, por ejemplo para los textos de San Agustín.
A principios del s. XX, desde el papado se recomendó el uso de esta pronunciación para todo el mundo. Sus rasgos distintivos, con respecto a la clásica, son:
Los diptongos ae, oe se pronuncian e (Ecclesiae: [ecclésie]).
Los grupos ce, ci, cae, coe, suenan palatales oclusivos sordos (Caecilia: [Chechília]), aunque en la Edad Media sonaba más bien [ts], asibilación, así como el grupo ti más vocal.
De manera análoga, los grupos ge, gi, gy , – gae, goe, muy escasos –, palatales oclusivos sonoros (legem: [léyem]; gymnasia: [yimnásia]).
El Grupo gn resulta una ñ española si no va a principio de palabra (agnus: [añus]; pero gnosis: [gnosis]).
La s sonora intervocálica y sorda inicial o tras consonante.
Los grupos sce, sci, scae, scoe, suenan palatales fricativos ([∫] ch en francés, sh en inglés; suscipe: [sú∫ipe]). Y los grupos xce, xci, xcae, xcoe, suenan sch (excelsis: [ekschelsis]).
where I have highlighted the bit we're interested in - and the example of Caecilia and how it's pronounced, to show that the author doesn't use the IPA, and in particular writes [ch] for the usual [ʧ].
As an aside, the penultimate listed feature (intervocalic -s- being always pronounced [z]) is supposedly inherited from neostandard Italian ("New Standard" Italian), which coexists with Standard Italian without threatening to replace it. For the record, Standard Italian has various exceptions to the voiced intervocalic sibilant, and Southern Italian has only the voiceless one, maintaining the Classical Latin pronunciation. This begs the question: did Ecclesiastical Latin really pick up this Northern Italian feature?
Phonologically, the "New Standard" picks up some Northern Italian contemporary peculiarities, as opposed to literary Tuscan-Florentine being the model for Standard Italian. There is a question about this on Italian.SE, here. Denis Nardin's answer quotes some excerpts in English from The Italian Language Today. The authors are Anna Laura Lepschy and Giulio Lepschy, both Northern Italian.
I just came across some old notes I took when reading Harold Copeman's Singing in Latin: Or Pronunciation Explor'd (unfortunately, I forget which edition it was).
In the section on Ecclesiastical/Italian Latin, Copeman says that xci/xce traditionally had /kʃ/ (I'm not sure exactly what that means, or on what basis he makes this statement*), but that "In Italian literary Latin xc is x-c, [kstʃ]. I understand that this is now generally preferred in singing; [ktʃ] is acceptable and [kʃ] is not recognized" (p.228). This agrees with what Vincenzo Oliva says in the comments and in his answer.
*As Rafael says in his answer, at least some editions of the Liber Usualis give [kʃ]. Here is a link to the English translation of the 1957 Liber Usualis, which gives [kʃ] as the pronunciation of "XC before the sounds of E and I" (p. xxij).