See
http://www.standardsforhighways.co.uk/dmrb/vol6/section2/td4295.pdf
from paragraph 7.25 onwards for guidance wrt widening on bends.

50m is considered a tight radius on a link if you are expecting artics or other large vehicles.

it is a v mountainous road where a 1m widening of the shoulder break point will result in the fill toe being a few hundred meters down the bank! not much option to straighten them out nor budget.

will read what you sent thanks

As Debaser said, whether to widen on curves will depend on your percentage of heavy vehicles.

How wide are the existing lanes? US research has shown that going from 3.3 to 3.6 m lanes has little effect on safety of two lane rural roads. You can expect a 5% reduction in lane departure (run-off-road and head-on) crashes. In terrain where widening usually means big fills or retaining walls, it might be more cost effective to put that money into other improvements.

The big payoff was widiening 2.7 or 3.0 m lanes to 3.3 m lanes.

http://www.tfhrc.gov/safety/pubs/99207.pdf, page 29

     "...students of traffic are beginning to realize the false economy of mechanically controlled traffic, and hand work by trained officers will again prevail." - Wm. Phelps Eno, ca. 1928

the current lanes are 3m and we are widening to 3.5m

This is purely an accademic question as there isn't the space t go much wider without spending lots of $$

How likely is it that two heavies will meet at one of these pinch points?

Probably worth doing a risk assessment and getting your view on paper, even if economics mean widening (and by extension reducing any hazard) can't be done. At least you will have said your piece, and be on the record (should you have concerns).

Will 'professional' drivers in your part of the world take notice of warning signs as a mitigation measure, e.g.
'Road Narrows'
http://webarchive.nationalarchives.gov.uk/+/http://www.dft.gov.uk/pgr/roads/tss/workingdrawings/warningp500series/517roadnarrowsononesideahead.pdf
and 'Oncoming vehicles in middle of road' type?
http://webarchive.nationalarchives.gov.uk/+/http://www.dft.gov.uk/pgr/roads/tss/workingdrawings/warningp500series/supplementaryplateoncomi4538.pdf
or maybe even yield the right of way?
http://webarchive.nationalarchives.gov.uk/+/http://www.dft.gov.uk/pgr/roads/tss/workingdrawings/regulatoryp600series/prioritytovehiclesfromth4358.pdf
http://webarchive.nationalarchives.gov.uk/+/http://www.dft.gov.uk/pgr/roads/tss/workingdrawings/regulatoryp600series/151givewaytooncomingvehicles.pdf

If you're otherwise providing a centreline along the road, it may actually be worth leaving it out over these tight radii and accompanying this measure with some sort of warning that large vehicles could be encroaching on your side of the road?

First, let me state that I have never designed a spiral transition curve.  However, I have done surveying work which has involved railroads and am well acquainted with the mathematics and evolution of spiral transition curves.

Question #1:  I don't know your particular geometry, but you should note that the length of spiral transition curves is not decided quantity, particularly in highway design.  In railroad design the length was more certain and determined based on the time rate change in acceleration.  This and a number of other methods have been used to determine the "recommended" length of easement curves.  The current AASHTO (2004) recommendation is based on the "recommended" super-elevation development length.  However, even the "recommended" superelevation development length is a matter of opinion.  Typically I still use the 1:200, but the 2004 AASHTO reduces this for lower speeds because it correlates it to tire friction.  What I am saying is that the curve radius and the transition curve and best super-elevation runoff are not necessarily directly related with one another and this is a changing area of code development.  You may not need to change the length of the spiral to meet the your current design code and provide a safe design.  

Question #2:  I wouldn't want to vary the lane widths, making wider tangents and narrower curves.  But then again I've never considered the proposition until now.  My initial reaction in lieu of additional information is a red flag.

I made a post about super-elevation a while back and no-one even responded.  So I hope this helps.

A bit late to respond however, there is an excellent college text " Route Surveys and design" by Hickerson and published by McGraw Hill.  This book explains in details Transition Spirals and Spiraled Compound Curves.