Hi Oliver,

If you can add a column, how about this first-aid?

| Product   |    g | kJ/100g |   kJ | kcal |
|-----------+------+---------+------+------|
| Bread     | 50.6 |    1372 |  694 |  166 |
| Butter    | 11.5 |    3054 |  351 |   84 |
| Marmalade | 19.7 |     926 |  182 |   44 |
|-----------+------+---------+------+------|
|           |      |         | 1227 |  294 |
#+TBLFM: $3='(org-lookup-first $1 '(remote(nf,@I$1..@II$1)) '(remote(nf,@I$2..@II$2)))
#+TBLFM: $4='(* $2 (/ $3 $b));N%.0f
#+TBLFM: $5=$4/$j;%.0f
#+TBLFM: @>$4..$5=vsum(@I..II)
(Each TBLFM line has no linebreak.)



From here, it is no more than a first-aid.
It's just a tracing log, and it has no conclusion for this issue.
I hope this would help you get the new direction, if you need.

You got the results calculated wrongly, that is:
> | Product   |    g |   kJ | kcal |
> |-----------+------+------+------|
> | Bread     | 50.6 |  694 |  166 |
> | Butter    | 11.5 |  158 |   38 |
> | Marmalade | 19.7 |  270 |   65 |
> |-----------+------+------+------|
> |           |      | 1122 |  269 |
> #+TBLFM: $3='(* $2 (/ (org-lookup-first $1 '(remote(nf,@I$1..@II$1)) '(remote(nf,@I$2..@II$2))) 100.0));N%.0f::$4=$3/$j;%.0f::@>$3..$4=vsum(@I..II)
This seems to me that:
#+BEGIN_SRC emacs-lisp
  (* 50.6 (/ 1372 100.0))                 ; => 694.2320000000001
  (* 11.5 (/ 1372 100.0))                 ; => 157.78
  (* 19.7 (/ 1372 100.0))                 ; => 270.284
#+END_SRC
You didn't want 1372 for all the cases, if I understood your calculation
correctly.

Where did it(=1372) come from?
Turning on the formula debugging with =C-c {=, and then Hitting =C-c *=
in the field of @2$3 (whose value is 694) said:
#+BEGIN_EXAMPLE