Well, apparently it's "just" folk etymology anyway. I didn't know; I found it in an interview with a professor of etymology (
http://www.hf.uio.no/ilos/for-ansatte/ILOS-posten/desember07/forskning.html – Norwegian only, I'm afraid!

) and assumed it was an accepted one. There's an excerpt of the alleged etymology on the bottom of the linked page, btw.
It takes time for new etymologies to establish themselves. Apparently we have a serious new proposal here, but will it carry?
How would this scholar explain that Old Danish/Swedish and Old Norwegian had älsk- with a half open vowel? Starting from **ailiska- one should expect **eilska in Icelandic/Norwegian and **elska in Old Swedish.
May be some conditioning?
And what is supposedly wrong with the traditional etymology < *aliska?
Well, I bought this dictionary and it is really the most up to date and profound dictonary on the core of the inherited Scandinavian vocabulary. Best of all, it has systematically treated the Indo-European parallels and consistently serves us with reconstructed forms, including Proto-Germanic, Paleo-Germanic, Pre-Germanic and PIE with laryngeals.
The author critisises the traditional established etymology for the Old Norse adjective
elskr (supposedly <
*aliska-) because it presupposes a derivate from a strong verb by means of
*-iska- for which there are no parallels. Also for the semantic development ('nourishing' > 'loving') he claims no parallels can be found.
He also believes it to be phonologically problematic that
*aliska- should have given
elskr.
*Daniska- has given
dansk 'danish' and
*batista- has given bazst (later by analogy
best) 'best'. He admits though on this third argument that several interpratations of the data is possible.
On
*ailisko:n he claims that *-aiCCC in this case could have given eCCC as in
helg 'holyday' (and not eiCCC as in Old Norse
beisk 'bitter'). He does not refer to the East Scandinavian data but a quick look at the words that he refers to as parallels seem to verify that in this environment before a heavy consonant cluster the /ä/ of Old Swedish
älska is not in contradicton with an origin from Common-Norse
*aiCC <
*aiCiC.
Upon reflection I see no reason why this new etymology could not become established as a first alterntive. NB though it is not a derivate from
eld/ild 'fire' but both words would be parallell derivates from the extinct stem
aila-, which is attested in Old Runic with a similar meaning as
eld/ild.