Next, why you choose the gloss "to subtract". Whats the proposed definition?
x1 is a difference/result of subtraction of x3 from x2 ?
Then the glossword should be "difference"
Bulgarian: разлика (bg) f
Czech: rozdíl (cs) m
Danish: differens (da) c
Dutch: verschil (nl) n
Esperanto: diferenco
Finnish: erotus (fi)
French: différence (fr) f
German: Differenz (de) f
Hebrew: הפרש (hefresh) m
Indonesian: selisih
Italian: differenza (it) f
Japanese: 差 (さ, sa)
Persian: اختلاف (fa) (extelâf), تفاضل (fa) (tafâzol)
Polish: różnica (pl) f
Romanian: diferență (ro) f
Russian: разность (ru) (ráznost’) f, разница (ru) (ráznitsa) f
Slovene: razlika (sl) f
Spanish: diferencia (es) f
Swedish: differens (sv) c, skillnad (sv) c
Telugu: భేదము (bhaedamu)
As for Mandarin it is cha1shu4
Of course we can play with the sounds a bit but in order to get a better mnemonic we have to choose either chinese or romance etymology.
So I suggest either {tcacu}/{tcaco} (.uinaisai it clashes with {tcaci}) or {difro} (final -o is normal for mekso gismu).
However, selsumji seems fine to me. But if you do really want to add it to jvs please consider that jian3 is not suitable.