Let's look at the docstring of
org-footnote-action (emphasis mine):
(org-footnote-action &optional SPECIAL)
Do the right thing for footnotes.
When at a footnote reference, jump to the definition.
When at a definition, jump to the references if they exist, offer
to create them otherwise.
When neither at definition or reference, create a new footnote,
interactively if possible.
With prefix arg SPECIAL, or when no footnote can be created,
offer additional commands in a menu.
So it seems that when point is at beginning of line it is considered no footnote can be created. Arguably, this could be made more explicit.
I actually wonder if it's a bug: the test is performed by function
org-footnote--allow-reference-p, where one can read on the source (for org-version 9.1.14):
(defun org-footnote--allow-reference-p ()
"Non-nil when a footnote reference can be inserted at point."
;; XXX: This is similar to `org-footnote-in-valid-context-p' but
;; more accurate and usually faster, except in some corner cases.
;; It may replace it after doing proper benchmarks as it would be
;; used in fontification.
As we can see, the beginning of line test is the very first one. Also, it turns out that the mentionned
org-footnode-in-valid-context-p function does return
t when point is at beginning of line.