Regardng the count mismatches, I suspect you may have multiple things going on here. I checked HMT-W4 not having BHS and it has 1465 occurrenecs of this word in various forms. I assume that is due to textual variation in the originals. Next ESVS, RSVS, NAS95S and KJVS all report different numbers of references to H1697. But I don't know what text each was actually translated from, and it could have been a collection of texts in the case of some translations. In order for everything to match properly I suspect you would need to have the Hebrew text from which the English was made and that text would have to be tagged 100% accurately.
As to the questions of mechanics let me see if I can help a little, though the following mostly repeats Ken's and the podcast's information. It depends upon where you start. Assuming that as you are using key numbers that you are starting in English you would find the key number from instant details when you mouseover on an English word, say in this case for example "words" in ESVS Gen 11:1. ID will show you the key and then you can run [KEY H1697] searches against any tagged English text you have. Once you have done that you can open an analysis tab which will give you a table of the translations found. Bear in mind that where a word in English is used to translate multiple Hebrew words it will show all those words. You just have to ignore the ones where the key does not match. Just looking quickly over this list here for the ESVS there is quite some variation. At that point I'd cross check with lexica. Decent lexica will give you a good appreciation of the range of meaning.
Regarding word@[key H1697]. That should, when the search is run against an English text, give you just where the word "word" is used in the translation (EXACT search) or variants of "word" such as "words", "worded" etc. (FLEX search). This search combine both the key and the translation you are looking for. Example output from the query against ESVS.
Total number of verses = 749
(total number of verses displayed = 749)
word@[key H1697] [Flex] (793 total words)
word = 444
H1697 dabar דָּבָר = 444
words = 349
H1697 dabar דָּבָר = 349
But all the above said, the way I normally try to get the range of meaning of a word in Greek (and would do it more or less similarly in Hebrew) is that I would use a lexicon to get the meanings as documented there. (You can do this even if you start from English by finding the Hebrew word you're interested in and using a a lexicon, or an expository dictionary). Then I would run a search for the Hebrew word in an original language text and then open various English translations in parallel panes and see how the word was handled in those translations. You tend to get more contextual information that way. You can see how the sentence was handled rather than just the word. But you can get an indication using key number searches.
So I'm pretty sure you can in fact do what you originally wanted to do, but if this is not what you are after perhaps you could walk through a detailed example highlighting where it goes awry for you.
Accordance Configurations :
Mac : 2009 27" iMac Windows : HP 4540s laptop
Intel Core Duo Intel i5 Ivy Bridge
12GB RAM 8GB RAM
Accordance 10.4.3.2 Accordance 10.4.3.2 and Aleph 10.4.3.2
OSX 10.9 (Mavericks) Win 7 Professional x64 SP1