I think that BMP code is a fine example of why using raw PICAXE variable names isn't as easy to fathom as using SYMBOL defined variable names.
Not being familiar with that sensor, what the code is doing, means I can't give any immediate help. The starting point might be to work through it, seeing what variables and @bptr addresses are used. Perhaps then converting some of the code to use named variables, seeing what's left, and rationalising that.
There do seem to be a lot of intermediate calculations required as shown in the datasheet but only a few bytes read from the sensor and a few bytes in the results so everything else except that which needs to be kept should be reusable.
If going the whole hog, I would probably start by renaming 'w' variables as 'xw' then symbol defining 'xw' = 'w', renaming 'b' variables as 'xw.msb' or 'xw.lsb' and adding symbols for those. Give 'xw' variables more meaningful names, then sort into order as to which ones are to be kept and which are reusable. Then allocate real 'w' variables as appropriate.