Maximal isometric contractions do not seem to be beneficial for PNF stretching in comparison to submaximal contractions. There is some inconclusive evidence that pre-stretch isometric contractions of around 2/3 of maximal intensity are optimal, but it might be that PNF is not actually better than plain static stretching.
There have been a few studies looking into the effect of different intensities of isometric contractions in PNF stretching, and they tend to find that there are only small differences in outcome with varying contraction intensity, with around 2/3 of maximal intensity being optimal, and with submaximal intensities also presenting a lower injury risk.
Feland & Marin 2004 compared contractions at 100%, 60%, and 20% of maximal, as well as a control group who did no stretching. They found that all treatment groups improved in flexibility compared to the control, but that there was no significant difference in flexibility gains between the treatment groups. They concluded that "CRPNF stretching using submaximal contractions is just as beneficial at improving hamstring flexibility as maximal contractions, and may reduce the risk of injury associated with PNF stretching."
Sheard & Paine 2010 ran a crossover study comparing contractions at 100%, 50%, and 20% of maximal. The crossover aspect means that each participant tried all three different contraction intensities, rather than being separated into groups that did one each. They found the greatest increases in ROM with 50% contractions, and through mathematical modelling were able to conclude that a contraction at 65% of maximal would have been optimal. Significant increases in ROM were seen with even just 20% contraction intensity. They concluded that "Although no evidence that maximal contraction intensity PNF actually causes injury is available in the literature, caution against the use of higher contraction intensities is becoming more common (3,4,9,15,24). On the basis of our findings, we recommend that when athletes and trainers perform PNF, they use a submaximal contraction intensity approaching 65% MVIC [(Maximal Voluntary Isometric Contraction)] to maximize ∆ROM while limiting the potential risk of injury that may accompany PNF application at, or approaching, 100% MVIC."
Kwak & Ryo 2015 compared contractions at 100%, 60%, and 20% of maximal, again with a control group who did no stretching, and found improvements in range of motion in all groups over the control, with the 100% and 60% groups seeing a slightly greater increase than the 20% group. However the statistical test they used (Duncan’s multiple range test) is considered excessively liberal is its potential for reporting false positives, so it may be safer to interpret their results as merely seeing that all the PNF groups saw increases in ROM over the control, but that the contraction intensity didn't actually make a significant difference. They concluded that "a moderate contraction intensity, e.g., 60%, may be optimal for healthy normal persons and athletes who need flexibility" while also suggesting that "those who feel difficulty in muscle contraction or muscle pain can sufficiently increase their ROM even with a contraction intensity of 20%".
However this does not mean that PNF stretching with a submaximal contraction is the best or most effective means of stretching. Unfortunately none of these studies used a 0% contraction intensity - i.e. plain static stretching with no pre-stretch contraction - as a control. The fact that contraction intensities as low as 20% were effective does leave me wondering whether the contraction is necessary at all.
A 2018 review by Lempke, Green, Murray & Stanek looked at research comparing PNF stretching to static stretching, and concluded that "the results show both static and PNF stretching effectively increase ROM; however, one does not appear to be more effective than the other."
In conclusion, if you want to increase your static flexibility, just stretch, don't worry too much about whether you should use static or PNF stretching, and if you do choose to use PNF, don't bother with maximal contractions.