To me, the North African (Western Arabic) forms look intermediate between those from the Gwalior temple and those used by Fibonacci. The Eastern Arabic forms don't. Fibonacci traveled back and forth to Algeria with his father in his childhood, where he studied mathematics. Although he credited the Indians with the method of calculating with place-value numerals, he never traveled to India, but he traveled extensively to North Africa.
Even if this were the case (I can't say that I can see such intermediarity clearly), we're not using Fibonacci's numeral shapes from Liber Abaci today so that doesn't seem to matter much. And as far as "we're using North African numbers today" is concerned, pretty much all of the European shapes used today that are similar to the North African ones from Fibonacci's time with the exception of 6 and 8 are the same shapes that I would be able to recognize in Gwalior inscriptions even if had no prior exposure to the history of numerals and you only told me to read those shapes as decimal numerals, so I'm not sure that one could say that these shapes were invented in North Africa, as opposed to just transferred through North Africa from India in a recognizable shape.
I only see minor differences between today's numeral shapes and the shapes on the pages of Liber Abaci I've been able to find. (Have you found a relatively complete online 13th-century edition?)
I agree that "transferred through North Africa from India in a recognizable shape" is pretty much the case—but not without change. As you point out, by looking at the 6 and 8, you can see that that's how the transmission happened. Even the 5, 7, and 9 seem recognizably intermediate to me.
