Daniel, re: your second reason, I’d argue that a DOTIMES loop would be less clear, as that list would probably be collected using the longer push-nreverse idiom:

(let (lst)
  (dotimes (i 5)
    (push i lst))
  (nreverse lst))

Otherwise, I agree with you.

1. I’ve never seen anyone write an iter-loop on one line like that. It’s always written with the accumulator directly below the driver. In this case, the extra parentheses are beside the point: it’s easier (for humans) to parse the code because the what-are-we-iterating-over and the what-are-we-doing parts are separate.
2. Comparing simple loops is uninformative. Both of those loops are so simple that you don’t need loop or iter. dotimes is plenty. The only proper test of an iteration construct is whether it can be used to write complicated loops effectively. After all, you don’t need any iteration constructs in scheme, as recursion is fine for relatively simple loops.

Your point that “clauses may appear arbitrarily deep in the body” is a good reason to prefer iterate for complex loops, but it is only for complex loops that you care about powerful iteration constructs.

(loop for i upto 5 collect i)

(iter (for i from 0 to 5) (collect i))

The LOOP example is obviously clearer for lack of parentheses.