I don't have the instrument, so I can't reproduce it completely, but I've done some research and if there is an error I don't think it's in the string conversion itself. I say this because:
From what I can gather, a typical output string would be the following: +4.27150000E-03
I took just a "scan from string" node and gave it that string as a string input and %.;%g; as the format specifier string. This yielded a numeric value of 0.0042715
So, as far as actually finding the root of the problem, there could be a few things going on. My first thought is some form corruption causing the data received to not be valid for conversion. There are two ways of testing this:
Open up a test panel in MAX and perform a loopback test, as described here: http://www.ni.com/white-paper/3450/en/#toc4
-or-
Install and open NI IO Trace and view the bus traffic as detailed here: http://digital.ni.com/public.nsf/allkb/282C5D41E2BA04F2862574BA007803B9
If the loopback tests works properly, go ahead and run the IO trace and post the results here.