Recumbent steering is odd. I don't think laying the head angle back is the way to go. Mike Burrows will advise a completely vertical head angle and reverse fork offset to give a little trail, for the optimum 'bent steering. See also the Ratcatcher SL
This theory has been proven/demonstrated as viable up to 100 mph by BMW, using motorcycles.
The issue with 'bent steering feeling 'squirrelly' is I believe down to wheel flop i.e. the degree to which the front end is lowered when the steering is turned. The slacker the head angle the greater is the tendency for the steering to flop sideways, to the extent that many recumbents just cannot be ridden hands-free other than perhaps by especially experienced and talented riders.....
I have seen some high-racer style recumbents where the head angle is slackened in order to reduce the tiller action of the steering. Users of such report that they are not evil handling. I think the importance of flop is overstated on a machine where the rider isn't moving his bodyweight in a gross fashion, and the smaller the trail value, the less it happens anyway, for small movements of the steering especially.
Mike has also experimented with many different rake and trail values on a single machine and IIRC he concluded that you would have a controllable machine of some kind over a wide range of settings, but that the steering feel would vary. One thing that you can't easily quantify is the possibility of shimmy; any castor will shimmy, it is just a question of when and how badly. So up to a point you can select your steering on the basis of preference, and then reject those settings which give genuinely unacceptable things like shimmy.
My take on the steering is that there are two things going on; trail steer and lean steer. Each has a particular time constant, with the reaction to trail steer usually being much faster than that of lean steer. Thus on a normal bike you get used to a particular combination of responses, and (in the simplest terms) you use lean steer to ride no-hands, and trail steer sorts out any minor perturbations and gives basic stability. The amount of lean and therefore lean steer input available is large, and the rider can alter the time period of the trail steer response just by sitting differently on the bike (affecting weight coupling), and/or holding the handlebars differently.
Any time you increase the coupled
weight over the front wheel, you alter the strength and possibly the time period of the trail steer response. On my recumbent machine the trail steer time period seems much shorter than normal until you sit upright, and then it changes (the weight is centred similarly but no longer coupled in the same way; that it changes at all makes nonsense of the importance of 'flop' IMHO ). By contrast the lean steer input available when recumbent is small, and the time period seems longer than normal. The result of all this is that the steering feels rather twitchy, but actually getting the blessed thing to lean-steer into a corner is kind of tricky.
In addition to everything else there is no real weight on the bars, and this affects both intentional and unintentional steering inputs. I have found that the steering behaves more 'normally' when I am pulling back on the bars, which may have a similar effect to leaning on the bars on an upright, because of the tiller action in both instances.
The normal things that you might do to the steering geometry to alter the lean-steer response can't be done without also affecting the trail-steer response, so it may be that there is no happy medium; that there are not more recumbents that steer nicely hands off is probably good evidence that this is the case.