I might just be putting puzzle pieces together that don't fit.... but hear me out anyway!
Like Cas posted, the daughter's name looks to me as some form of Louise.
You mentioned that Selina was born in Kent, to mmn Cowan.
Could this be her birth:
SMITH, SELINA
MARGARET
JANE COWEN
GRO Reference: 1886 D Quarter in CANTERBURY Volume 02A Page 793
There is an electoral record for George Allan and Selina
J S Allan at 11
Montagu Terrace in 1925.
Could George have given the names and mixed up her middle initials with the daughter, as there is electoral records for:
Louise
J S Allan and Selina
M J Allan living at
Junction Street, Great (Leith), in the 1936/7.
They appear at other places as well.
If the above is right, then George would have died (according to the newspaper, clip attached) in 1932, and Selina's mother, Jane, must have remarried to a Hodgson.
There is this death at SP:
HODGSON JANE age 77 mmn Cowan, 1926, 685 / 1 / 11 St Bernard
And a possible for the Louise, if she didn't marry:
ALLAN Louise
Jane
S age 73 1988 473 / 386 Stirling
(Could Louise have been a daughter born to Selina before she married? Or after the 1921 census? I can't find a birth for her under Smith or Allan in England or Scotland)
This clipping has all the right addresses and names, but the wrong year for George's death? Is this just an almost "duplicate family"? Clip is from The Falkirk Herald and Scottish Midlands Journal dated 11 February 1933
Edit: looking at the reference number for the 1921 census for Falkirk, I note that Jane Hodgson, George Allan, and Selina Allan are all on that page, but no Louise or variation.