Why is what's popular popular in general? The biggest obstacles to popularity for Lisp, as I see it (and I'm not an expert) are the prefix syntax and its attendant parentheses; the history of having been chiefly used for AI, which is rather esoteric; and the (undeserved) reputation for slowness. Paul Graham has more to say in http://www.paulgraham.com/iflisp.html . None of these are intrinsically bad, but they are often perceived as such.
The interviewers are surprised because Lisp isn't popular, as you observed, and nobody expects it to be used for anything big.
Is Lisp impractical? Not really. As I understand it, the one problem it has is lack of libraries. Why? Because it's not popular. Even so, most things you want to do are probably librarified already, as "lack" here is relative.
Is Arc impractical? Not in the abstract, but it sounds like it will be for you. I don't think that Arc should be the first language you pick up, for a few reasons. It is still in a state of flux; it is missing certain features; it is designed for experience hackers; and, most importantly, there is a real paucity of documentation. http://arcfn.com/ has some great stuff, but that's about it. Starting in Arc would, therefore, be difficult.
Learning another Lisp would be a good idea, however; I would recommend learning Scheme. You might try working through How To Design Programs (http://www.htdp.org/) in Dr. Scheme (http://www.plt-scheme.org/); as I learned Scheme as a later language, I never worked through it, but from what I see and what I hear, it seems like a very good text. (I've never used Common Lisp, so I don't feel qualified to talk about it, but if someone wants to put in a good word for it, that might be helpful.) After you get a handle on that, you would almost certainly be able to pick up Arc in its undocumented state.
And if you really want a challenge, of course, you could always jump right into Arc; sometimes that is the best way to learn something.
Personally, when learning Lisp, syntax wasn't a big issue. While the infix syntax of most languages is fairly intuitive, the rest of syntax (of e.g. C++ or Perl) is scary compared to Lisp's parens. (Although perhaps what puts people off is the frightening simplicity of Lisp's lack of syntax after using a conventional language.)
One the things that proved difficult about learning Lisp was choosing an implementation. There is no canonical implementation and as such I had to try several out before I learned which ones I enjoyed using. This survey of CL implementations (http://common-lisp.net/~dlw/LispSurvey.html) helped me when deciding which implementations to try. I recommend both CLISP (http://clisp.cons.org/) and SBCL (http://www.sbcl.org/), but that is just personal preference.
Also, the IDE was a difficult issue. I eventually settled on emacs with slime (http://common-lisp.net/project/slime/), although I have occasionally used Cusp, a Lisp plugin for Eclipse (http://bitfauna.com/projects/cusp/index.html). The main problem here was a lack of (thourough, easy to follow, up to date) instructions. Over the last couple of years I have gradually found good instructions on slime, but I can't seem to find any of them right now...
I wasn't saying that the prefix syntax was an issue, just that it hampered popularity. Perception ≠ reality, after all. I too quite like the prefix syntax. I still slightly miss the simplicity of some of the Ruby code that I wrote, but the prefix syntax always wins me over; its benefits (macros) outweigh the minor downsides. Especially with []s and ssyntax.
And thanks for the CL thoughts! Regarding IDEs, I myself have ended up working just from the Mac OS X text editor TextMate (using Visor, which puts Terminal on a hotkey), and haven't found it problematic.
I'll second the recommendation for Scheme in general and PLT Scheme in particular. Arc is built on top of PLT Scheme at the moment. Another great book is The Structure and Interpretation of Computer Programs (http://mitpress.mit.edu/sicp/).
