We've done some of both. We did the blood test, but it was for a MASSIVE panel of things. They did one draw and tested her for: fish, shellfish, dairy, soy, peanuts, rice, wheat, cockroach, cat, dog (these were environmental, not food!), almonds, brazil nuts, cashews, pistachio, sesame seeds, walnuts, pecans, eggs...a few more things I can't remember.
We did one set of scratch tests, for about 5 things. I found it a little...fuzzy. Dd actually reacted to the negative control; they did it twice, once with a brand new bottle of stuff, to make sure it wasn't a contamination issue...apparently, she has sensitive skin. She came back completely negative to soy via skin prick (it was smaller than her negative control) and had very low numbers on her RAST, but seemed to have a very, very mild reaction--a bit of diaper rash--when we trialed it after the scratch test. So, in that case, I think the RAST was more accurate. For things to which a child is highly allergic, I think skin test is fine--dd's scratch for rye blew up, and since we know she has a severe wheat allergy, that made sense. For more borderline things, I think it can be hard to interpret.
We'll do the blood testing again next summer (a year later) and I plan to throw in everything and the kitchen sink: all her allergens, plus some common things for which she hasn't been tested (like chocolate) and hasn't had or has had very mild reaction to (like tomatoes). If they're doing the draw, they should test for as much as humanly possibly.
If you suspect allergies in your daughter, I wouldn't limit it just to the foods you're allergic to--the propensity for allergies is hereditary, but the specific allergies are not (thus many people with seasonal allergies, like my dh, have food allergic children). You're the consumer. Your doctor works for you. Insist on the tests you want or fire him. FYI, they usually keep the blood for a few days and you can order more tests after the original round--we didn't initially test dd for tree nuts, but when she came back + for peanuts, my doctor called the lab and had them run all the other nuts with the sample they already had.
From what I've heard, positive results on the blood draw are around 50% accurate, negative results are about 90% accurate. So it's definitely not ideal, but as you do more tests over time and likely have a set of real-world reactions to go off, you can get a decent picture.