I agree with many things in your post, but I think that one of your important claims is wrong, and that weakens your argument significantly and shows a fundamental misunderstanding of Intel's situation:
At the time no one was asking for server class ARM chips (which is basically what XScale was).
XScale was not a server class ARM chip. StrongARM was designed from the beginning for the embedded market. XScale was being used in BlackBerry and Palm handheld devices at the time Intel got rid of XScale. (Example from 2005. There are many more examples if you want to google around.)
So, Intel was already in the mobile game, and it decided to quit it right as it was taking off. That fact remains regardless off your points about stacks and modems. Intel could have made the latter without issue. (In fact, it's making modems now.)
Intel's decision to focus rather than branch out was shortsighted and clearly cost them. That said, "focusing" is a classic business strategy, so I can't blame them too much for following it. However, I think that focusing doesn't pay off the way that it used to. Many of the most successful tech companies at the moment (Google, Amazon, etc.) are branching out like mad, and it's paying dividends.