Long dated treasuries might add no more to or detract no more from portfolio rewards as that of stock. Ponder
this link - where generally the differences are just 'noise' i.e. half in stocks or 25% in each of stock and long dated treasuries broadly yielded similar rewards.
A short and long (20 year) dated barbell of treasuries overall compares to a 10 year bullet. As does a rolling 10 year ladder. With a ladder where each bond is held to maturity then there's no need to mark to market, each years reward is approximately the average of the current and past nine years 10 year yield to maturities. As of last year that was down to around 2% levels and for 2021 it will earn 1.74%. 10% of bonds maturing each year, perhaps 5% of total portfolio value, that will roll into higher (or lower) 10 year maturity yields each year, or that might perhaps buy twice as many shares as a year earlier if stocks drop a lot in price.
For the likes of a Permanent Portfolio, if you hold 25% stock, 50% in a 10 year ladder that's not market to market, 25% gold - then historically that's provided a 4.5% annualised real (modest gains) since 1896 with very few negative years and only very marginally negative when such years did occur (very low risk).
As for holding government bonds right now, it is difficult to see the justification for retail investors.
As you say, not easy to drop high street fixed term bonds into a stocks/shares ISA. Also any amounts 'deposited' are fully protected, even £millions - the state would rather print money or increase taxes rather than allow a default to occur.
A Permanent Portfolio rewarding perhaps 4.5% annualised real with very low year on year downside risks and that holds 50% in Gilts ... or perhaps more in a all-stock alternative that might make more, but could turn around and bite back hard (large potential declines), and IMO writing off bonds/safety in order to speculate isn't perhaps the wisest of choices. What might hit longer dated bonds hard, could hit stocks as equally as hard, if not even harder. Assuming long dated gilts to have only one way to go is comparable to saying the same about stocks.